Untouched open Tcl bug tickets: Not tested

Triage of 818 open and Pending Bug tickets of Tcl with no comment or check-in from me on 2026-10-16 (in Tcl's workflow Pending means "resolved, waiting for confirmation"). Reproduced on Linux (main 9.1 tip, 9.0 tip, 8.7b1, 8.6 tip, 8.5, 8.4 and old releases), on FreeBSD, NetBSD, OpenBSD and OpenIndiana VMs for platform tickets, and on a Windows 11 VM. Rows marked patch link to a prototype patch (287 in all, in tcl-triage/prototypes). Updated 2026-11-11: the emoji before each ticket id shows its progress.

Needs Windows 4

⬜e5c21e566f2025inconsistent handling of glob -nocomplainpatchNEEDS-WINDOWS (unresolved: the SMB ENOENT error does not occur against a Windows 11 SMB server; needs the reporter's server type); Windows EACCES c…

Open ticket e5c21e566f · 37. File System · filed 2025-08-13

🧪 Prototype check (2026-10-24)

e5c21e566f.patch: NOT-REPRODUCED-ON-TIP (the reported SMB case cannot be triggered here; the patch builds and is harmless)

  • Base main (tip c092dae4af)
  • Before the SMB "no such file or directory" error does not occur against the Windows 11 SMB server: e5c.tcl gives empty lists for a missing local dir, //localhost/C$/<missing>, a missing share and a deep missing path, both as admin and as basic user. New with the basic-user token: a dir with a DENY(RD) ACE gives couldn't read directory ".../priv/*.ext": permission denied (EACCES) for priv, priv/. and priv/sub, local and via \\localhost\C$. That is consistent on Windows, unlike the Unix stat/opendir difference; as admin it is listed anyway
  • Patch applied cleanly, 7 changed lines, not changed
  • After all cases unchanged (the EACCES case is deliberately not covered by the patch); file tests admin 8 failed, low 6 failed, as baseline
  • Test none possible (needs an SMB server that returns ENOENT-class codes other than FILE/PATH_NOT_FOUND)
Full verdict

NEEDS-WINDOWS (unresolved: the SMB ENOENT error does not occur against a Windows 11 SMB server; needs the reporter's server type); Windows EACCES case not testable as the elevated ssh admin (a DENY list-directory ACE does not stop listing)

  • Platform generic (EACCES part) + Windows (ENOENT on SMB part)
  • Tested Linux, dir priv mode 000: glob -nocomplain -directory priv * → "couldn't read directory "priv/": permission denied"; -directory priv/. and priv/sub → empty, no error (main, 9.0.5, 8.6.18, 8.5.0, 8.6.0); 8.4.20 → empty for all.
  • Cause unix/tclUnixFile.c TclpMatchInDirectory: a failed stat() of the directory (ENOENT or EACCES) is silently "no match", but a failed opendir() is an error, so the result depends on which of the two calls fails. Windows (win/tclWinFile.c TclpMatchInDirectory): the directory passes GetFileAttributesW, then FindFirstFileExW fails with a code not in IsNoSuchFileError() (only FILE_NOT_FOUND, PATH_NOT_FOUND, INVALID_NAME) but which Tcl_WinConvertError maps to ENOENT (e.g. ERROR_NO_MORE_FILES, ERROR_BAD_NETPATH, ERROR_BAD_NET_NAME, which SMB servers return) → the reported "no such file or directory" error, with the pattern appended to the "directory" name.
  • Prototype prototypes/e5c21e566f.patch (32 lines, Windows only, syntax-checked with mingw, not run; no test — needs SMB): treat any FindFirstFileExW error that maps to ENOENT as "no match". EACCES part: none (design question; dgp wants real errors reported, so the fix would be the opposite direction or a new option needing a TIP).
  • Suggested ticket comment On Windows the error comes from FindFirstFileExW() returning a code that maps to ENOENT but is not in IsNoSuchFileError() (likely ERROR_NO_MORE_FILES/ERROR_BAD_NETPATH from SMB). Proposed fix: treat every ENOENT-class error as "no match" (patch available, needs testing on an SMB share). The EACCES inconsistency is because stat() failure is silent but opendir() failure is an error.
Windows VM, run 4
  • Platform generic (EACCES part, see result-41.md) + Windows (ENOENT on SMB)
  • Verdict NEEDS-WINDOWS (unresolved: the SMB ENOENT error does not occur against a Windows 11 SMB server; needs the reporter's server type); Windows EACCES case not testable as the elevated ssh admin (a DENY list-directory ACE does not stop listing)
  • Tested VM work-win4/e5c.tcl on 9.1b1 and 8.6.18: glob -nocomplain -directory <d> *.ext returns an empty list without error for a missing local dir, //localhost/C$/<missing>, //localhost/C$/no/such/deep/path, //localhost/nosuchshare/x, and a dir with a DENY(RD) ACE (listed anyway). Related: for paths longer than MAX_PATH 9.x glob silently returns nothing (see 3eace1b21f).
  • Cause (from Linux analysis) win/tclWinFile.c TclpMatchInDirectory treats only FILE_NOT_FOUND/PATH_NOT_FOUND/INVALID_NAME as "no match"; other ENOENT-class codes from SMB servers become errors
  • Prototype prototypes/e5c21e566f.patch (existing, Windows only) now compiled with MSVC 2022 on tcl-wt91: builds cleanly; the cases above unchanged; fileName, fileSystem, fCmd, cmdAH 0 failed, winFCmd 3 failures as unpatched. The SMB case itself remains untested.
  • Suggested ticket comment Against a Windows 11 SMB server (\\localhost\C$, missing dirs, missing share) glob -nocomplain returns an empty list on 8.6.18 and 9.1. Which SMB server gave the error? The proposed patch (treat every ENOENT-class FindFirstFileExW error as "no match") compiles and passes the file tests on Windows.
⬜184358681b2017file readable/writable does not work with Windows Offline FilesNEEDS-WINDOWS (unresolved: Offline Files needs the CSC service and an offline-available share, i.e. system configuration changes not allowed on thi…

Open ticket 184358681b · 16. Commands A-H · filed 2017-06-13

Full verdict

NEEDS-WINDOWS (unresolved: Offline Files needs the CSC service and an offline-available share, i.e. system configuration changes not allowed on this VM; CscService is stopped/manual)

  • Platform Windows
  • Tested code reading of win/tclWinFile.c:NativeAccess on main/9.0/8.6; Unix is unaffected (access(2)). Repro: win/184358681b.tcl <file-on-offline-share> ?<dir>? (needs a share made "Always available offline", synced, then Work offline).
  • Cause in 8.6.6, NativeAccess decided R/W access only via GetFileSecurityW + AccessCheck, which fails for the Client-Side-Caching copy when offline. Since branch fix-1613456fff, a regular file is first checked with CreateFileW(GENERIC_READ/WRITE), which should work offline; directories still use GetFileSecurityW/AccessCheck and may still report 0.
  • Prototype none (needs a Windows Offline Files setup to confirm first)
  • Suggested ticket comment Since 8.6.9 (92c4bfbe32) file readable/writable open the file with CreateFileW instead of only checking the ACL, which should cover offline files; directories still use the ACL check. Can you retest with a current release?
Windows VM, run 4
  • Platform Windows
  • Verdict NEEDS-WINDOWS (unresolved: Offline Files needs the CSC service and an offline-available share, i.e. system configuration changes not allowed on this VM; CscService is stopped/manual)
  • Tested not run (win/184358681b.tcl needs a synced offline share)
  • Cause see result-36.md (8.6.9+ NativeAccess opens regular files with CreateFileW; directories still use the ACL check)
  • Prototype none
  • Suggested ticket comment (as in result-36.md) Since 8.6.9 (92c4bfbe32) file readable/writable open the file with CreateFileW instead of only checking the ACL, which should cover offline files; directories still use the ACL check. Can you retest with a current release?
⬜2913585fff2009[win32] serial channel blocks indefinitely on writeNEEDS-WINDOWS (unresolved: the VM has no serial port — all four VirtualBox UARTs are off and no com0com is installed; adding one needs a VM configu…

Open ticket 2913585fff · 25. Channel System · filed 2009-12-13

Full verdict

NEEDS-WINDOWS (unresolved: the VM has no serial port — all four VirtualBox UARTs are off and no com0com is installed; adding one needs a VM configuration change while the user is running it)

  • Platform Windows
  • Tested code reading: win/tclWinSerial.c on main still sets WriteTotalTimeoutConstant/Multiplier to 0 and SerialBlockingWrite waits in GetOverlappedResult(..., TRUE); open.n still says -timeout does not affect writes. Unix not checked (no serial hardware; a pty has no flow control).
  • Cause (by reading) SerialBlockingWrite waits forever for a write that the flow control never lets through; the writer thread then keeps the handle open after [close].
  • Prototype none (needs the Windows repro first; possible fix is to forward a write timeout to SetCommTimeouts)
  • Repro win/2913585fff.tcl ?COMn? — needs a COM port with no peer (VirtualBox serial port in "Disconnected"/host pipe mode with no reader, or com0com with the other side closed). It runs the write+close in a child with -handshake rtscts, blocking 0 and 1, and reports HANG / REOPEN-FAILED; prints a VERDICT line and exits by itself.
  • Suggested ticket comment (after the Windows run) Still present on 9.x: a pending serial write with no flow control never finishes, and [close] then hangs (non-blocking) or leaves the port open (blocking).
Windows VM, run 2
  • Platform Windows
  • Verdict NEEDS-WINDOWS (unresolved: the VM has no serial port — all four VirtualBox UARTs are off and no com0com is installed; adding one needs a VM configuration change while the user is running it)
  • Tested nothing new (GetPortNames is empty on the VM)
  • Cause (by reading, see the Linux analysis) win/tclWinSerial.c sets zero write timeouts and SerialBlockingWrite waits in GetOverlappedResult(..., TRUE)
  • Prototype none
  • Suggested ticket comment none yet; to test, power the VM off, enable UART1 in "Disconnected" mode (or a host pipe with no reader) and run win/2913585fff.tcl COM1.
⬜2049971fff2008clock command doesn't reflect Windows time zone changesNEEDS-WINDOWS (by code reading, still present in main, 9.0, 8.6)

Open ticket 2049971fff · 06. Time Measurement · filed 2008-08-13

Full verdict

NEEDS-WINDOWS (by code reading, still present in main, 9.0, 8.6)

  • Platform Windows
  • Tested read generic/tclClock.c and library/clock.tcl on main; on Linux checked that touching env(TZ) clears the cached system zone after the 1 s refresh
  • Cause the system zone is cached (9.0+: ClockGetSystemTimeZone keyed on TzsetIfNecessary's epoch; 8.6: ::tcl::clock::CachedSystemTimeZone). On Windows without TZ, the epoch changes only when TCL_TZ/TZ changes, so a registry time zone change is never noticed. The 2008 workaround (unset CachedSystemTimeZone) no longer works in 9.0. The 9.0 workaround: set env(TZ) :UTC; clock format 0; unset env(TZ).
  • Prototype none (needs a WM_SETTINGCHANGE/registry-change hook or a periodic GuessWindowsTimeZone; a design decision)
  • Repro win/2049971fff.tcl (changes the system zone with tzutil and restores it; needs admin, so ask before running it on a VM someone uses)
  • Suggested ticket comment Still cached in 9.0, now in the C engine. The old workaround (unset ::tcl::clock::CachedSystemTimeZone) no longer applies; touching env(TZ) works instead. This should become an RFE, or be fixed by checking the registry zone when the 1 s TZ refresh fires.

Cannot test 47

✅c35e347c6f2026Concurrent Tcl_Init calls race in Unix locale initialization, causing crashes on OpenBSDCANNOT-TEST (OpenBSD) — does not reproduce on Linux/glibc (expected); Serhiy reproduced it on OpenBSD 7.9 amd64 (see comments). The original Tcl_In…

Open ticket c35e347c6f · 20. [interp] · filed 2026-10-07

✅ Since triage (updated 2026-11-11)

Status now: Closed (Fixed)

  • 🔧 2026-11-05 check-in 14d675e4b4 on unix-locale-init-lock, in main: Merge trunk
  • 💬 2026-10-18 jan.nijtmans: Question: Is this already fixed now by 63941fee4f? Maybe this patch is no longer necessary.
  • 💬 2026-10-18 serhiy.storchaka: Partly. Since 63941fee4f, Tcl_Init() no longer calls setlocale() once the initial encoding is set, so tcl-thread-init-race.tcl no longer crashes on OpenBSD (0 of 10 runs, 10 of 10 just before that merge). But Tcl_GetEncodingNameFromEnvironment() still calls setlocale() without a lock: from [encoding user], from Tcl_GetEncodingNameForUser(), and from every Tcl_Init() as long as the initial encoding was not found.…
  • 💬 2026-10-19 gustafn3: Jan. By looking at the referenced change, I expect it to prevent the crash in the original reproducer: once initial encoding setup succeeds in the main thread, subsequent interpreter initializations no longer enter the locale-detection path. However, the change does not synchronize the setlocale() / nl_langinfo() sequence itself. The initialEncodingSet flag is unsynchronized, and direct calls to Tcl_GetEncodingNam…
  • 💬 2026-10-19 jan.nijtmans: Understood. Just one small remark: variable 'ds' is shadowed. Fixed 6e63a280be @ashok, can you have a final look?
  • 💬 2026-10-20 apnadkarni: Not looked at this yet, but generally speaking, my preference is to keep interp creation (Tcl_Init being part of it) as performant as possible. Part of that is minimizing calls as well as synchronization. To that end, is there a reason for retrieving the locale from the environment on a per interp basis and not once per process (or per thread)? If not, could this be subsumed into Tcl_InitSubsystems? Thoughts? …
  • 💬 2026-10-20 gustafn3: Ashok. yes, process-wide setup would be preferable. From the embedding application's perspective, there is no need to reset the process locale whenever another interpreter is initialized. The expected performance differences depend on the initialization scope: - Per-interpreter setup repeats locale detection for every interpreter. - Per-thread setup reduces that work when a thread creates multiple interpreters, bu…
  • 💬 2026-10-20 serhiy.storchaka: Since 63941fee4f it already works mostly that way: the locale is read once, by TclpSetInitialEncodings() called from Tcl_FindExecutable() after Tcl_InitSubsystems(), and its call in Tcl_Init() returns at once as soon as the initial encoding is set. So the lock in unix-locale-init-lock is not taken when an interp is created, except while the initial encoding could not be set yet. It is needed for the public Tcl_G…
  • 💬 2026-10-20 gustafn3: Thanks, Serhiy ... this resolves the performance aspect. With the successful-initialization guard in place, the locale lock is outside the normal interpreter-creation path, so the concern about repeated synchronization is already addressed there. The combination makes sense to me: avoid repeated initialization once it has succeeded, and protect the locale operations when explicitly reached through the public query…
  • 💬 2026-11-06 apnadkarni: Fixed for 9.0 in 4f47985ff9 and 9.1 in 40f9131022 respectively. Thanks for the fix.
Full verdict

CANNOT-TEST (OpenBSD) — does not reproduce on Linux/glibc (expected); Serhiy reproduced it on OpenBSD 7.9 amd64 (see comments). The original Tcl_Init() trigger is gone on trunk/9.0 since the fix for 63941fee4f (00b1026efe trunk, 27f849b6b3 9.0, 2026-10-17, unreleased); [encoding user] / Tcl_GetEncodingNameForUser() still call setlocale() unlocked

  • Platform unix (OpenBSD; glibc and macOS lock setlocale() internally)
  • Tested Linux, Thread 3.0.7: c35e347c6f-encoding-user-race.tcl 10/10 OK on main (2026-10-22) and 10/10 OK on branch tip 743a63c60b; tcl-thread-init-race.tcl parallel 8 50 OK on the branch; encoding.test, unixInit.test 0 failed on the branch
  • Cause unix/tclUnixInit.c Tcl_GetEncodingNameFromEnvironment(): setlocale(LC_CTYPE, "") + nl_langinfo(CODESET) without a lock
  • Existing work branch unix-locale-init-lock (tip 743a63c60b, 2026-10-19, rebased by Jan; the old pre-rebase tip 424917fbce is closed), unmerged (not in trunk/9.0/8.6); merges into current trunk without conflicts (dry run); waiting for Ashok's review
  • Prototype none (branch exists)
  • To run on OpenBSD build main and the branch with Thread, run tclsh c35e347c6f-encoding-user-race.tcl 10x each (main should crash with a double free in free(), the branch should not)
  • Suggested ticket comment none needed (discussion is current).
⬜15d60c0d2e2025make fails for tk8.6.16 on macOSCANNOT-TEST (macOS). This looks like the user's environment, plus a side bug in tcltk-man2html that reproduces on Linux.

Open ticket 15d60c0d2e · 69. Other · filed 2025-07-24

Full verdict

CANNOT-TEST (macOS). This looks like the user's environment, plus a side bug in tcltk-man2html that reproduces on Linux.

  • Platform macOS (build)
  • Tested code reading of tk8.6 unix/Makefile.in, configure.in and tcl.m4; the side issue on Linux with tcltk-man2html.tcl from 8.6 and 9.0. The ticket's attachments (config.log, make install output) are missing from the repository: "cannot resolve name".
  • Cause --enable-framework adds html-tk to make install. BUILD_HTML runs $(BUILD_TCLSH) = ${TCL_BIN_DIR}/tclsh. With --with-tcl=/usr/local/tcl8.6.16/lib (an install dir) no tclsh exists there, so it falls back to TCL_EXE (SC_PROG_TCLSH = the first tclsh* on $PATH). On this machine that is an old 8.6.6 tclsh, with no TCL_LIBRARY set, and it finds /usr/local/lib/tcl8.6/init.tcl from a newer install: "have 8.6.6, need exactly 8.6.16". Tk itself is installed; only the HTML docs step is affected. Side bug: tools/tcltk-man2html.tcl takes the *newest* tk* sibling of --srcdir (lsort ... end in 8.6, findversion in 9.x). That is why the 8.6.16 install says "using Tk source directory tk9.0.2" and builds 9.0.2 HTML docs. Reproduced on Linux with sibling tk8.6.18 and tk9.0.5 trees: both the 8.6 and the 9.0 scripts pick tk9.0.5.
  • Prototype none. The main failure is environmental. The man2html part needs Tk's Makefile to pass --useversion (or the script to prefer the srcdir it is run from), which is a Tk/Tcl Makefile change to decide on.
  • Suggested ticket comment The error comes from the HTML docs step of a framework install. With --with-tcl pointing at an install dir there is no build tclsh, so make uses the first tclsh on PATH, here an old 8.6.6 one. Workaround: make install TCL_EXE=/usr/local/tcl8.6.16/bin/tclsh8.6 (or point --with-tcl at the Tcl build dir). Separately, tcltk-man2html.tcl picks the newest tk* directory next to the source tree, which is why the tk9.0.2 docs were built.
⬜ba4a2704d02025make fails for tk9.0.2 on unixCANNOT-TEST (macOS). Same cause as 15d60c0d2e (duplicate).

Open ticket ba4a2704d0 · 69. Other · filed 2025-07-24

Full verdict

CANNOT-TEST (macOS). Same cause as 15d60c0d2e (duplicate).

  • Platform macOS (build)
  • Tested code reading (as for 15d60c0d2e). The error text again names an 8.6.6 tclsh (init.tcl search path /usr/tcl8.6.6/library) while installing Tk 9.0.2.
  • Cause the framework make install runs html-tk through TCL_EXE = the first tclsh on PATH (an old 8.6.6), because --with-tcl=/usr/local/tcl9.0.2/lib has no build tclsh. Not a Tk build failure as such.
  • Prototype none
  • Suggested ticket comment Same cause as 15d60c0d2e: the HTML docs step of the framework install runs an old 8.6.6 tclsh found on PATH. Use make install TCL_EXE=/usr/local/tcl9.0.2/bin/tclsh9.0, or configure Tk with --with-tcl pointing at the Tcl build directory. Suggest closing as a duplicate of 15d60c0d2e.
⬜57f27dbe1a2024Sporadic failure in httpPipeline-3.9h-normalCANNOT-TEST (needs the external test server test-tcl-http.kerlin.org, constraint serverNeeded; no outside network allowed)

Open ticket 57f27dbe1a · 29. http Package · filed 2024-05-12

Full verdict

CANNOT-TEST (needs the external test server test-tcl-http.kerlin.org, constraint serverNeeded; no outside network allowed)

  • Platform generic (tests)
  • Tested test code reading only (tests/httpPipeline.test sections 4–5, RunTest)
  • Cause unknown. The pattern (tokens 2..9 all got error "X" right after the request was sent, then all overlapped) looks like the persistent connection being closed by the server/at EOF while the pipelined queue was being written. That is related to, but not necessarily the same as, 05b158ed56 (a reused socket closed between geturl and OpenSocket); unverified.
  • Prototype none
  • Suggested ticket comment Cannot be run without the kerlin.org test server. If the main session wants to pursue it, a local httpd with 1 s keep-alive timeout and the ?delay/close query semantics (section 4 of httpPipeline.test) would allow running httpPipeline-3.9h repeatedly (N>=40) with and without the 05b158ed56 prototype.
⬜25dddef8182024MacOSX GNUmakefile does not use settings providedCANNOT-TEST (macOS)

Open ticket 25dddef818 · 53. Configuration and Build Tools · filed 2024-04-16

Full verdict

CANNOT-TEST (macOS)

  • Platform macOS
  • Tested code reading on main. macosx/GNUmakefile runs ../unix/configure itself with fixed --enable-framework --enable-dtrace --disable-zipfs plus ${CONFIGURE_ARGS} ${EXTRA_CONFIGURE_ARGS}; options given to a separate macosx/configure run are not used by it (the GNUmakefile is meant to be driven via make EXTRA_CONFIGURE_ARGS=..., see macosx/README). The "unrecognized options framework/dtrace" warnings in sub-makefiles come from unix/configure.ac forwarding all of its arguments to bundled packages via PKG_CFG_ARGS (same mechanism as 17ffc49790) - harmless warnings.
  • Cause design of macosx/GNUmakefile (configure args are make variables, not taken from a prior configure run).
  • Prototype none (macOS-only, needs design; no macOS here)
  • Suggested ticket comment The GNUmakefile takes configure options through make EXTRA_CONFIGURE_ARGS="..." rather than from a separate macosx/configure run; the framework/dtrace warnings come from passing Tcl's configure arguments on to the bundled packages and are harmless. Would documenting EXTRA_CONFIGURE_ARGS more prominently be enough?
⬜1e8185c7fb2023www.tcl-lang.org pages in need of updateCANNOT-TEST (no network; the website source is not in any local repository)

Open ticket 1e8185c7fb · 69. Other · filed 2023-12-24

Full verdict

CANNOT-TEST (no network; the website source is not in any local repository)

  • Platform site (www.tcl-lang.org, not in tcl.fossil)
  • Tested nothing. aku said on 2024-01-08 that the home page and /doc/ items are fixed. The rest (compile howto, platforms page, about/language.html, ActiveState links) needs someone with website access.
  • Cause n/a
  • Prototype none
  • Suggested ticket comment (for whoever maintains www) The home-page and /doc/ items were reported fixed. Can you check the remaining pages: compile.html, platforms.html, about/language.html and the ActiveState links? Then close or split this ticket.
⬜bbe7c6ff9e2023zipfs failures on MacOS with password protected zip filespatchCANNOT-TEST (macOS) — still present: tests/zipfs.test on main still skips the password tests on Darwin (constraint bbe7c6ff9e)

Open ticket bbe7c6ff9e · - New Builtin Commands · filed 2023-09-21

🧪 Prototype check (2026-10-24)

bbe7c6ff9e.patch: VERIFIED (mechanism only, under emulation; macOS root cause unconfirmed)

  • Base main (tip d6dd6d0f71)
  • Before macOS not available. Emulated "a libz whose get_crc_table() is not the classic table" with LD_PRELOAD work-a2/v2/crctab_shim.c (returns byte-swapped entries): make test TESTFLAGS="-file zipfs.test -match *password*" -> 10 failed, all invalid password. Without the shim the tip passes on Linux. tests/zipfs.test on the tip still disables these tests on Darwin.
  • Patch applied with offsets +6 / +34 (re-saved, 11 changed lines, content unchanged)
  • After same shim run -> 29 passed, 0 failed (tclZipfs.c no longer calls get_crc_table); full make test TESTFLAGS="-file zipfs.test": 535 total, 498 passed, 37 skipped, 0 failed; builds without warnings
  • Test the re-enabled password tests are the test; they fail before / pass after only under the emulation, not on Linux natively. CAUTION: the emulated failure is "invalid password", while macOS CI reported "invalid CRC" / "decompression error" (the password check passed there), so the get_crc_table hypothesis is still unconfirmed; it needs a macOS CI run.
Full verdict

CANNOT-TEST (macOS) — still present: tests/zipfs.test on main still skips the password tests on Darwin (constraint bbe7c6ff9e)

  • Platform macOS
  • Tested Linux main (passes); code inspection
  • Cause (hypothesis): generic/tclZipfs.c is always compiled against the bundled compat/zlib headers (Makefile: -I$(ZLIB_DIR)), but on macOS it is linked with the system libz; the traditional PKWARE decryption keys use crc32tab = get_crc_table(), and only the password paths (init_keys/zdecode) use that table, which matches exactly the failing tests (wrong keys → garbage → "invalid CRC"/"decompression error"). Apple's libz need not return a standard 256-entry table there.
  • Prototype prototypes/bbe7c6ff9e.patch (44 lines, speculative; zipfs.test passes on Linux): use the static crc_table from the bundled crc32.h that tclZipfs.c already includes instead of get_crc_table(), and re-enable the tests on Darwin so a CI run on macOS (core-* tag) validates it.
  • Suggested ticket comment Only the decryption keys use get_crc_table(), and tclZipfs.c is compiled with the bundled zlib headers but linked with the system libz on macOS. Proposed (untested on macOS) fix: use the bundled crc_table directly; the patch re-enables the tests so macOS CI can confirm.
🔧112af76b092023Double-checked locking pervades the Tcl implemetationCANNOT-TEST (memory-ordering race, not observable on x86-64; confirmed by code reading: Tcl_MutexLock()/Tcl_ConditionWait2() on main read the lazil…

Open ticket 112af76b09 · 41. Memory Allocation · filed 2023-05-21

🔧 Since triage (updated 2026-11-11)

Status now: Open

  • 🔧 2026-10-14 check-in fed33b2ff4 on double-checked-lock-ordering, not merged: Fix 112af76b09: use an acquire load for the unlocked check and a release store for publishing in the lazy initialization of mutexes and condition variables (T
Full verdict

CANNOT-TEST (memory-ordering race, not observable on x86-64; confirmed by code reading: Tcl_MutexLock()/Tcl_ConditionWait2() on main read the lazily created pointer without a lock and without acquire ordering)

  • Platform generic (relevant on weakly ordered CPUs: ARM, POWER; no effect on x86)
  • Tested branch tip fed33b2ff4 built on Linux x86-64 with Thread 3.0.7: thread.test, unixNotfy.test, notify.test, event.test, socket.test all 0 failed; reviewed the diff (84 changed lines: acquire/release macros in tclInt.h via __atomic_* / MSVC ReadAcquire/WriteRelease, used in unix/tclUnixThrd.c, unix/tclUnixNotfy.c, win/tclWinThrd.c, win/tclWinPipe.c, win/tclWinSerial.c; the pipe/serial flags are now stored after the initialization, which is correct)
  • Cause unix/tclUnixThrd.c Tcl_MutexLock/Tcl_ConditionWait2 (and the Windows twins, notifier start, pipe/serial init): unlocked check of a pointer/flag published by a plain store
  • Existing work branch double-checked-lock-ordering (tip fed33b2ff4, 2026-10-14), unmerged (not in trunk/9.0/8.6); merges into current trunk without conflicts (dry run). The API part of the ticket (Tcl_Once, explicit init) would need a TIP
  • Prototype none (branch exists)
  • Suggested ticket comment none needed (Serhiy's 2026-10-14 comment is current). Testing on ARM64 (e.g. the OpenBSD arm64 VM or Windows on ARM) would be the only way to exercise it, and even there a failure is unlikely to show.
⬜8d71525c182023expect 5.45.4 + tcl 8.6.13 segfault on macOS 13 (Ventura)CANNOT-TEST (macOS 13 on ARM)

Open ticket 8d71525c18 · 45. Parsing and Eval · filed 2023-02-17

Full verdict

CANNOT-TEST (macOS 13 on ARM)

  • Platform macOS
  • Tested code reading of unix/tclUnixPipe.c on 8.6 tip
  • Cause crash is inside dyld's _dyld_fork_child atfork handler in the child after fork() in TclpCreateProcess, i.e. in the OS/expect environment, not in Tcl code. Since 8.6.14 (42d3f0cdb9, 2023-07-26, fix for c54e4a1aeb) TclpCreateProcess prefers vfork() when available, which macOS has, so the fork()+atfork path from the backtrace is no longer taken by Tcl's exec.
  • Prototype none
  • Suggested ticket comment The crash is in dyld's fork-child handler, outside Tcl. Since 8.6.14 &#91;exec&#93; uses vfork() instead of fork() where available (see c54e4a1aeb), so this path should be gone; can someone on macOS 13+ ARM confirm with expect built against 8.6.14 or later? Otherwise suggest closing.
⬜62f56a61ac2022Mac OSX build of Tcl aborts on startupCANNOT-TEST (macOS) — needs a non-framework ./configure && make shell on macOS without macher; unix/Makefile.in still appends the zip to the dyli…

Open ticket 62f56a61ac · 53. Configuration and Build Tools · filed 2022-04-20

Full verdict

CANNOT-TEST (macOS) — needs a non-framework ./configure && make shell on macOS without macher; unix/Makefile.in still appends the zip to the dylib with cat when macher is not found, so it is probably still the case (code reading on main 2026-10-22)

  • Platform macOS (build)
  • Tested code reading only
  • Cause zipfs data appended to libtcl9.x.dylib invalidates the Mach-O code signature (macOS kills the process: "Killed: 9")
  • Existing work none (Serhiy asked on 2026-10-14 for someone with a Mac to check)
  • Prototype none (macOS)
  • Suggested ticket comment none (the 2026-10-14 question is still open).
⬜055ed8476a2021load-10.1 failsCANNOT-TEST (8.5 only — 8.5 is no longer maintained; needs a host with a noexec temp dir, which requires root to set up)

Open ticket 055ed8476a · 37. File System · filed 2021-12-13

Full verdict

CANNOT-TEST (8.5 only — 8.5 is no longer maintained; needs a host with a noexec temp dir, which requires root to set up)

  • Platform unix (CentOS 7, core-8-5-branch only)
  • Tested code comparison only: 8.5 and 8.6 both copy the library to a temporary directory before dlopen
  • Cause "failed to map segment from shared object: Operation not permitted" is what dlopen reports for a file on a noexec mount, i.e. /tmp was mounted noexec on that machine; 8.5 and 8.6 use the same temp-file logic, so the 8.6 pass was probably an environment difference (TMPDIR) rather than a code fix.
  • Prototype none
  • Suggested ticket comment The error is dlopen() refusing a file on a noexec /tmp; 8.5 and 8.6 use the same temp location logic (TMPDIR first). As 8.5 is no longer maintained, propose closing as Out of Date.
⬜c837eb353e202164 bit build of TCL coredumps in AIXCANNOT-TEST (AIX/xlc)

Open ticket c837eb353e · 47. Bytecode Compiler · filed 2021-03-17

Full verdict

CANNOT-TEST (AIX/xlc)

  • Platform unix (AIX 7.1, xlc 64-bit)
  • Tested none (no AIX). Reading: JUMP_PEEPHOLE_F/JUMP_PEEPHOLE_V in generic/tclExecute.c are unchanged in structure on main/9.0/8.6; the macros are plain pointer arithmetic on pc, gcc/clang/MSVC are fine. The reporter's workaround (disable the peephole for _AIX) points to an xlc optimizer/codegen bug rather than a Tcl bug; the reported CFLAGS was -g only, so it may also be the computed-goto-free switch in xlc.
  • Cause unknown; probably compiler (xlc -q64) miscompiling TEBCresume.
  • Prototype none (cannot verify; reporter's one-line #if !defined(TCL_COMPILE_DEBUG) && !defined(_AIX) is the obvious candidate if someone with AIX confirms on 8.6.18/9.0)
  • Suggested ticket comment Cannot test without AIX. Could you retry with 8.6.18 or 9.0.4, and say which xlc version? If it still crashes, does it also crash with gcc on AIX? That would tell whether a !defined(_AIX) guard (or rather an xlc version check) is the right fix.
⬜7cd721e9ea2020mkdtemp() unavailable on Solaris 2.8patchCANNOT-TEST (Solaris 8; illumos has mkdtemp() in stdlib.h). Prototype verified on illumos with the fallback forced.

Open ticket 7cd721e9ea · 37. File System · filed 2020-06-19

🧪 Prototype check (2026-10-24)

7cd721e9ea.patch: VERIFIED

  • Base main (tip d6dd6d0f71)
  • Before emulated a libc without mkdtemp on Linux: out-of-tree build of the tip (work-a2/v2/nomk) with CFLAGS="-g -Dmkdtemp=no_such_mkdtemp" -> linking tclsh fails, undefined reference to «no_such_mkdtemp» (the tip's configure.ac has no mkdtemp check)
  • Patch applied cleanly (82 changed lines incl. regenerated configure, 69 hand-edited, unchanged); ~/tcltk/autoconf-2.73/bin/autoconf -f regenerates a configure byte-identical to the patched one
  • After same CFLAGS plus ac_cv_func_mkdtemp=no (work-a2/v2/nomk2): builds, compat/mkdtemp.c compiled without warnings, libtcl exports no_such_mkdtemp; file tempdir -> /tmp/tcl_XXXXXX mode 0o40700; file tempdir <dir>/tdx works; error path can't create temporary directory: no such file or directory; cmdAH + fCmd tests: 17306 total, 17231 passed, 75 skipped, 0 failed. Normal in-tree build: libtcl uses glibc mkdtemp@GLIBC_2.2.5 as before
  • Test none possible (build fallback; exercised only on systems without mkdtemp); the emulated build above stands in
Full verdict

CANNOT-TEST (Solaris 8; illumos has mkdtemp() in stdlib.h). Prototype verified on illumos with the fallback forced.

  • Platform build
  • Tested main (30d8ccce48) on Linux, configured with ac_cv_func_mkdtemp=no in a separate build dir to exercise the fallback: libtcl defines its own mkdtemp, file tempdir works (dir created, mode 0700, error path OK), cmdAH.test passes (17001/0 failed). Normal build still uses glibc's mkdtemp.
  • Cause missing compat fallback for mkdtemp(), which only became POSIX in 2008.
  • Prototype prototypes/7cd721e9ea.patch (69 hand-edited lines, 62 of them the new compat/mkdtemp.c modeled on compat/mkstemp.c; plus AC_REPLACE_FUNCS(mkdtemp ...), Makefile.in rule, tclConfig.h.in, declaration in tclUnixPort.h; regenerated configure with autoconf 2.73). No test (fallback only builds on systems lacking mkdtemp).
  • Suggested ticket comment Proposed fix: a compat/mkdtemp.c used via AC_REPLACE_FUNCS, like compat/mkstemp.c (mkdir() with EEXIST retry is race-free). Tested on Linux by forcing ac_cv_func_mkdtemp=no; needs someone with Solaris 8 to confirm, if that platform still matters.
Unix VMs, run 1
  • Platform build
  • Verdict CANNOT-TEST (Solaris 8; illumos has mkdtemp() in stdlib.h). Prototype verified on illumos with the fallback forced.
  • Tested OpenIndiana, gcc 14.3, prototypes/7cd721e9ea.patch applied to main (out-of-tree builds). Normal configure: "checking for mkdtemp... yes", libtcl uses the libc mkdtemp (UNDEF). With ac_cv_func_mkdtemp=no: compat/mkdtemp.c is compiled without warnings and libtcl defines mkdtemp. In both builds file tempdir creates /var/tmp/tcl_XXXXXX with mode 0700, file tempdir /tmp/tdx works, the error path says "can't create temporary directory: no such file or directory", and cmdAH.test passes (17001 total, 0 failed).
  • Cause unix/tclUnixFCmd.c:TclpCreateTemporaryDirectory uses mkdtemp() and configure does not check for it.
  • Prototype prototypes/7cd721e9ea.patch (69 hand-edited lines; unchanged). Note: like compat/mkstemp.c, the fallback exports a global mkdtemp symbol from libtcl.
  • Suggested ticket comment Proposed fix: compat/mkdtemp.c via AC_REPLACE_FUNCS, like mkstemp. Verified on OpenIndiana with the fallback forced (ac_cv_func_mkdtemp=no): file tempdir and cmdAH.test work. Solaris 8 itself was not tested.
⬜5d5c6dbded2020Broken links on www.tcl.tkCANNOT-TEST (no network; website source not available locally)

Open ticket 5d5c6dbded · 69. Other · filed 2020-05-16

Full verdict

CANNOT-TEST (no network; website source not available locally)

  • Platform site (www.tcl.tk / www.tcl-lang.org)
  • Tested nothing. The report dates from 2020 and lists mostly conference-page and ActiveState links. Overlaps with 1e8185c7fb.
  • Cause n/a
  • Prototype none
  • Suggested ticket comment (for the website maintainer) Many of these are old conference pages and dead third-party links. Can someone with website access re-run a link checker and close this ticket, or merge it with 1e8185c7fb?
⬜e67618079f2020Win rdp forwarded serial port: setting -sysbuffer gives "can't setup comm buffers: invalid argument"CANNOT-TEST (needs a serial port forwarded over RDP to Windows)

Open ticket e67618079f · 27. Channel Types · filed 2020-04-22

Full verdict

CANNOT-TEST (needs a serial port forwarded over RDP to Windows)

  • Platform Windows
  • Tested code reading only (win/tclWinSerial.c SerialSetOptionProc, unchanged since 2020 apart from error conversion).
  • Cause -sysbuffer calls SetupComm() and then GetCommState/SetCommState() to adjust XonLim/XoffLim; the RDP serial redirector driver rejects SetupComm sizes (ERROR_INVALID_PARAMETER) and SetCommState (ERROR_DEV_NOT_EXIST / "no such device"), as it also does for -handshake. Driver limitation rather than a Tcl bug. Jan's attempt 8b444e9115 (branch bug-e67618079, 2020-04-22) did not help per the reporter and was never merged.
  • Prototype none (no hardware; driver-specific)
  • Suggested ticket comment The errors come straight from SetupComm()/SetCommState() of the RDP serial redirector, which seems not to support these calls. Branch bug-e67618079 did not help and is unmerged; unless someone with such a setup can test, this is probably a driver limitation.
✅852a17eeb92019Build test fCmd.test failedCANNOT-TEST (the VM's page file is at the default location, so the failure does not occur; reproducing needs the page file moved off C:\ or disable…

Open ticket 852a17eeb9 · - New Builtin Commands · filed 2019-11-27

✅ Since triage (updated 2026-11-11)

Status now: Closed (Fixed)

  • 🔧 2026-10-29 check-in e373a1adc9 on fCmd-test-constraints, in main: Rebase fCmd-test-constraints to main. Fix [219158], 300724b80c and 852a17eeb9: fCmd.test constraints (read-only directory rename, the lost notWine definitio
  • 🔧 2026-11-10 check-in 7aa93727ce on fCmd-test-constraints, in main: Rebase Fix [219158], 300724b80c and 852a17eeb9: fCmd.test constraits
  • 🔧 2026-11-11 check-in f9dfe7796a on trunk, in main, 9.0: Fix [219158], 300724b80c and 852a17eeb9: fCmd.test constraits
  • 🔧 2026-11-11 check-in 8068a3d2b3 on core-9-0-branch, in 9.0: Fix [219158], 300724b80c and 852a17eeb9: fCmd.test constraits
  • 💬 2026-11-11 jan.nijtmans: Fixed in f9dfe7796a and core-9-0-branch.
Full verdict

CANNOT-TEST (the VM's page file is at the default location, so the failure does not occur; reproducing needs the page file moved off C:\ or disabled, a system setting I was told not to change)

  • Platform Windows
  • Tested code reading of tests/fCmd.test on main (2026-10-22); repro script win/852a17eeb9.tcl (prints ExistingPageFiles, the fCmd-30.3 result and a VERDICT line; optional arg: Tcl source dir to run fCmd.test -match fCmd-30.3*). Note: a VM with a page file on C: will just say PASSES; the failure needs no page file or one on another drive
  • Cause tests/fCmd.test fCmd-30.3 assumes C:\pagefile.sys
  • Existing work branch fCmd-test-constraints (tip ed6326c9e3, 2026-10-12), unmerged; takes the page files from the registry (ExistingPageFiles) and adds constraint pagefile; conflicts with current trunk in tests/fCmd.test (see 219158ffff)
  • Prototype none (branch exists)
  • Suggested ticket comment none needed until the branch is rebased.
Windows VM, run 6
  • Platform Windows
  • Verdict CANNOT-TEST (the VM's page file is at the default location, so the failure does not occur; reproducing needs the page file moved off C:\ or disabled, a system setting I was told not to change)
  • Tested read-only queries: Win32_PageFileUsage = C:\pagefile.sys (1280 MB, AutomaticManagedPagefile = True), registry ExistingPageFiles = \??\C:\pagefile.sys. win/852a17eeb9.tcl with tclsh91: "fCmd-30.3 PASSES here"; fCmd.test -match fCmd-30.3* on 9.1: Passed 1, Failed 0. (The 9.0/8.6 uninstalled builds could not load the registry package from library/, so the script ran on 9.1 only; the test assumption is the same everywhere.)
  • Cause tests/fCmd.test fCmd-30.3 hard-codes $env(SystemDrive)/pagefile.sys (code reading, see result-6x).
  • Prototype none (branch fCmd-test-constraints exists, unmerged; not built here since the VM cannot show the difference)
  • Suggested ticket comment none.
⬜07d9a953042018"tdbc::mysql configure" fails with MySQL 8.0.3CANNOT-TEST (no tdbcmysql sources/repository and no MySQL server available offline)

Open ticket 07d9a95304 · 81. Bundled Packages · filed 2018-10-11

Full verdict

CANNOT-TEST (no tdbcmysql sources/repository and no MySQL server available offline)

  • Platform generic (bundled package tdbc::mysql)
  • Tested none; the tdbcmysql sources are not in the local repositories (pkgs/ in the Tcl trees is empty, there is no tdbcmysql.fossil) and there is no mysqld/mariadbd installed
  • Cause tdbcmysql.c reads the session variable @@TX_ISOLATION in ConnectionConfigure, which MySQL 8.0.3 removed (the replacement is @@TRANSACTION_ISOLATION, available since 5.7.20)
  • Prototype none (package lives in the separate tdbcmysql repository)
  • Suggested ticket comment This belongs to the tdbcmysql repository. Someone with a MySQL 8 server should check whether the current tdbcmysql release still queries @@tx_isolation; if so, the fix is to use @@transaction_isolation when the server version is 5.7.20 or newer.
⬜35a1bb18732017Test Failure - socket-14.0.0 + socket-14.2 on MacosCANNOT-TEST (macOS)

Open ticket 35a1bb1873 · 69. Other · filed 2017-11-07

Full verdict

CANNOT-TEST (macOS)

  • Platform macOS
  • Tested Linux main: socket.test passes completely (run together with the event tests above); no macOS machine. The 2017 failures (empty -error after the async connect to an IPv4-only server via "localhost", and no writable event on refused connect) were seen on High Sierra with 8.7 at 84de05583b.
  • Cause unknown; likely the macOS (kqueue/CoreFoundation) notifier or getaddrinfo ordering ::1 before 127.0.0.1 for localhost in the async connect fallback (unix/tclUnixSock.c TcpConnect). The tests now carry localhost_v4 constraints.
  • Prototype none
  • Suggested ticket comment Ticket is from 2017 (8.7a1 era). Does make test TESTFLAGS="-file socket.test -match socket-14.*" still fail on a current macOS with 9.0/main? If not, this can be closed.
⏳e50d66e6c32017Crash on exit after loading dll built by mingw g++CANNOT-TEST (no 32-bit toolchain on the VM: C:\msys64\mingw32 contains only empty bin/include/lib skeleton dirs, no gcc/g++; the only g++ is ucrt64…

Open ticket e50d66e6c3 · 17. Commands I-L · filed 2017-02-16 · Pending (resolved, waiting for confirmation)

⏳ Since triage (updated 2026-11-11)

Status now: Pending (Works For Me)

    Full verdict

    CANNOT-TEST (no 32-bit toolchain on the VM: C:\msys64\mingw32 contains only empty bin/include/lib skeleton dirs, no gcc/g++; the only g++ is ucrt64 x86_64 16.2.0, and no 32-bit tclsh is built either; installing mingw-w64-i686-gcc was not allowed)

    • Platform Windows (32-bit MinGW only)
    • Tested Linux only as a smoke test of the repro (g++ .so with a static C++ object + exceptions, no Init function, loaded into main and 8.6.18, also with TCL_FINALIZE_ON_EXIT=1): clean. No 32-bit (i686) MinGW cross compiler here.
    • Cause unknown; likely 32-bit-only: 32-bit MinGW g++ used DWARF-2 unwinding (libgcc_s_dw2-1.dll, __register_frame_info/__deregister_frame_info), and "requested the Runtime to terminate it in an unusual way" is abort() from that runtime when it is torn down in the wrong order at process exit. 64-bit MinGW uses SEH, which would explain why 64-bit does not reproduce.
    • Existing work none (Pending, question to reporter 2026-10-12)
    • Prototype none
    • Repro win/e50d66e6c3.tcl + win/e50d66e6c3/{cppext.cpp,build.sh}: build cppext32.dll in an MSYS2 mingw32 shell (CXX=i686-w64-mingw32-g++ ./build.sh) and run with a 32-bit tclsh; also cppext64.dll from mingw64 for 64-bit. Three variants (load without Init, with TCL_FINALIZE_ON_EXIT=1, with an init name), prints VERDICT.
    • Suggested ticket comment none until a 32-bit run is done; if 32-bit also stays clean, close as Works For Me.
    Windows VM, run 6
    • Platform Windows (32-bit MinGW only)
    • Verdict CANNOT-TEST (no 32-bit toolchain on the VM: C:\msys64\mingw32 contains only empty bin/include/lib skeleton dirs, no gcc/g++; the only g++ is ucrt64 x86_64 16.2.0, and no 32-bit tclsh is built either; installing mingw-w64-i686-gcc was not allowed)
    • Tested nothing new (Serhiy's earlier 64-bit runs with 8.6.18 and 9.1 stay the only data; 64-bit uses SEH, not the DWARF-2 unwinder suspected here)
    • Cause unknown (suspected 32-bit libgcc DWARF-2 unwinder teardown order at exit, see result-6x)
    • Prototype none
    • Suggested ticket comment none until a 32-bit run is possible (needs pacman -S mingw-w64-i686-gcc in MSYS2 plus a 32-bit Tcl build, e.g. nmake MACHINE=IX86 or the mingw32 configure build).
    ⬜0b39decc762017clock.test 50.2 fails for negative values in IRIX: broken mktime()CANNOT-TEST (IRIX/MIPSPro)

    Open ticket 0b39decc76 · 06. Time Measurement · filed 2017-01-11

    Full verdict

    CANNOT-TEST (IRIX/MIPSPro)

    • Platform unix (IRIX)
    • Tested none; code read only. main still calls mktime() for :localtime scans (generic/tclClock.c ConvertLocalToUTCUsingC)
    • Cause IRIX libc mktime() returns -1 for all pre-epoch times; platform bug, not Tcl's.
    • Prototype none (obsolete platform; a compat mktime would be needed)
    • Suggested ticket comment This is an IRIX libc limitation (mktime() rejects pre-1970 values). IRIX is no longer a supported platform; suggest closing as Won't Fix.
    ⬜bf029ced862017binary.test 40.3 signed NaN fails on IRIX/MIPS machinespatchCANNOT-TEST (MIPS/RISC-V hardware; no qemu-riscv64 here) — test still expects 1 -NaN* on main, 9.0 and 8.6; x86-64 passes

    Open ticket bf029ced86 · 48. Number Handling · filed 2017-01-11

    🧪 Prototype check (2026-10-24)

    bf029ced86.patch: VERIFIED (on x86-64 and by pattern check; the MIPS/RISC-V run is still CANNOT-TEST)

    • Base main (tip d6dd6d0f71)
    • Before x86-64 cannot show the failure (binary scan \xFF\xFF\xFF\xFF f1 a -> -NaN(7ffffe0000000), sign kept); no qemu-riscv64/qemu-mips here. Checked the patterns instead (work-a4/v4nan.tcl): the old glob 1 -NaN* rejects 1 NaN(7fffffe0000000) and 1 NaN, the reported MIPS/RISC-V results
    • Patch applied cleanly (4 changed lines)
    • After new regexp ^1 -?NaN accepts 1 -NaN(...), 1 NaN(...), 1 NaN and still rejects 1 1.0; binary.test 754 total, 673 passed, 81 skipped, 0 failed on x86-64
    • Test the patch is the test fix; cannot fail on available hardware
    Full verdict

    CANNOT-TEST (MIPS/RISC-V hardware; no qemu-riscv64 here) — test still expects 1 -NaN* on main, 9.0 and 8.6; x86-64 passes

    • Platform generic (test portability; affects MIPS R4400/R10000 and RISC-V, per the 2026-10-13 comment)
    • Tested read binary.test on main/9.0/8.6; ran the patched binary.test on main (x86-64)
    • Cause generic/tclBinary.c:ScanNumber returns Tcl_NewDoubleObj(fvalue); the float→double widening may drop the sign of a NaN (RISC-V fcvt.d.s always gives the canonical +NaN), and IEEE 754 leaves that sign unspecified.
    • Prototype prototypes/bf029ced86.patch (4 changed lines, test binary-40.3 now -match regexp -result {^1 -?NaN} plus a comment); binary.test passes on x86-64. (A code alternative, building the double bits by hand for NaN floats, would also keep the sign and payload, but seems unnecessary.)
    • Suggested ticket comment The sign of a NaN after float→double conversion is not defined (RISC-V always gives +NaN). Proposed: binary-40.3 accepts both NaN and -NaN.
    ⬜40ef293eed2016Path intrep invalidation triggered by [cd] too aggressive?CANNOT-TEST (needs trofs, not available offline). Core behaviour unchanged.

    Open ticket 40ef293eed · 37. File System · filed 2016-04-26

    Full verdict

    CANNOT-TEST (needs trofs, not available offline). Core behaviour unchanged.

    • Platform generic (VFS)
    • Tested code reading on main (30d8ccce48) and 8.6: generic/tclIOUtil.c:Tcl_FSChdir still calls Tcl_FSMountsChanged(NULL) when the new cwd is in a different filesystem than the old one (since 48beab9ffd, 2015-06-23). That bumps theFilesystemEpoch and invalidates every cached path intrep; Tcl_FSConvertToPathType/TclFSEpochOk then re-resolve paths, which re-enters the filesystem's pathInFilesystem proc.
    • Cause the crash was infinite recursion in trofs (PathInFilesystem -> FetchDirFromCache -> Tcl_FSGetInternalRep -> PathInFilesystem); dgp's 2016 patch to trofs fixes it there. The question left open is a design one: whether a cross-filesystem cd needs to invalidate everything.
    • Prototype none (design question; the crash fix belongs to trofs)
    • Suggested ticket comment Core still invalidates all path intreps on a cd that crosses filesystems (48beab9ffd); the crash itself was fixed on the trofs side by the patch below. Unless someone wants to make the invalidation per-filesystem (the unused fsPtr argument of Tcl_FSMountsChanged), this could be closed or turned into an RFE.
    🔧40f628e8e32015Tcl_ListObjReplace() callers fail to handle errors (LIST_MAX) properlyCANNOT-TEST (the remaining oo::define case needs objc > LIST_MAX: unreachable on 64-bit 9.x; on 8.6 it needs two expanded lists of 268M elements, ~…

    Open ticket 40f628e8e3 · 14. List Object · filed 2015-11-19

    🔧 Since triage (updated 2026-11-11)

    Status now: Open

    • 🔧 2026-10-14 check-in fa198b47b0 on oo-define-list-overflow, not merged: Fix 40f628e8e3 in oo::define: check the result of Tcl_ListObjReplace() in MagicDefinitionInvoke(), and pass the real length of the list to Tcl_EvalObjv().
    Full verdict

    CANNOT-TEST (the remaining oo::define case needs objc > LIST_MAX: unreachable on 64-bit 9.x; on 8.6 it needs two expanded lists of 268M elements, ~2 GB each plus a 4+ GB eval stack, not run on the shared machine). The 2015 linsert/concat cases were fixed by dgp in 2015. Code confirmed unchanged: "TODO: overflow?" unchecked Tcl_ListObjReplace() in MagicDefinitionInvoke() on main, 9.0 and 8.6.

    • Platform generic
    • Tested main own build with work-a3/oo62.tcl (oo::define C {*}[lrepeat 50000000 x] under ulimit -v): allocation failures panic ("list creation failed: unable to alloc"), before the unchecked error path can be reached, so no crash repro. Branch tip: oo.test, ooNext2.test, ooProp.test pass; oo62.tcl with 1000 args OK.
    • Cause generic/tclOODefineCmds.c MagicDefinitionInvoke(): if Tcl_ListObjReplace() failed, Tcl_EvalObjv() was called with objc - cmdIndex words from a 1-element list (read past the end).
    • Existing work oo-define-list-overflow (fa198b47b0, 2026-10-14, 16 changed lines, on main), unmerged. Applies cleanly to main; needs a manual port to 9.0 (1 of 2 hunks fails) and 8.6 (both hunks fail; 8.6 has /* TODO: overflow? */ at a different place).
    • Prototype none (branch exists)
    • Suggested ticket comment none needed (Serhiy's 2026-10-14 comment points to the branch).
    ⬜4db0832c5b2015Source build AIX with xlc make test -> Segmentation faultCANNOT-TEST (AIX/xlc not available)

    Open ticket 4db0832c5b · 53. Configuration and Build Tools · filed 2015-06-24

    Full verdict

    CANNOT-TEST (AIX/xlc not available)

    • Platform build (AIX 7.1, xlc 12.1, OBJECT_MODE=64, Tcl 8.6.4)
    • Tested none; report has no backtrace or failing test name
    • Cause unknown
    • Prototype none
    • Suggested ticket comment There is no AIX system to test on, and the report has no failing test or backtrace. If it still happens with 8.6.18 or 9.0, please run make test TESTFLAGS="-verbose bpe" to find the test and attach a dbx backtrace; otherwise suggest closing.
    ⬜e3da32d8162015max os x [glob] returns the decomposed filenamesCANNOT-TEST (macOS) — no normalization code in glob/filesystem paths on main (unix/, macosx/, tclFileName.c, tclIOUtil.c), so behaviour is unchanged

    Open ticket e3da32d816 · 44. UTF-8 Strings · filed 2015-04-21

    Full verdict

    CANNOT-TEST (macOS) — no normalization code in glob/filesystem paths on main (unix/, macosx/, tclFileName.c, tclIOUtil.c), so behaviour is unchanged

    • Platform macOS
    • Tested code reading on main; checked the new 9.1 command
    • Cause HFS+ stores names in NFD and glob returns them as is. APFS (macOS 10.13+) keeps names in the form they were created with, so the reporter's script is likely to print "equal" on current macOS.
    • Prototype none (macOS; leaving glob as is is the intended design, per dkf)
    • Suggested ticket comment Tcl 9.1 adds &#91;unicode tonfc&#93; (TIP 726), so <code>unicode tonfc &#91;glob ...&#93;</code> is the iconv workaround in pure Tcl; glob itself is meant to keep returning the names as stored. Suggest closing once someone confirms on 9.1 / APFS.
    ⬜17a26ddf662014fCmd-27.5 and 27.6 fail because permission to set attribute is deniedpatchCANNOT-TEST (needs a file whose group cannot be re-set by the owner, e.g. inherited from a setgid dir of a foreign group on NFS/idmapped storage; o…

    Open ticket 17a26ddf66 · 37. File System · filed 2014-10-22

    🧪 Prototype check (2026-10-24)

    17a26ddf66.patch: VERIFIED

    • Base main (tip d6dd6d0f71)
    • Before simulated the reporter's environment with an LD_PRELOAD shim (work-a2/v2/chgrp_shim.c: chown() with gid != -1 returns EPERM): LD_PRELOAD=chgrp_shim.so ./tcltest ../tests/fCmd.test -match 'fCmd-27.[56]' -> both FAILED, could not set group for file "foo.tmp": operation not permitted
    • Patch applied with offset -9 (re-saved, 6 changed lines, content unchanged)
    • After same command -> 2 passed, 0 failed; full fCmd.test without shim: 305 total, 260 passed, 45 skipped, 0 failed
    • Test the patch is itself the test change; fCmd-27.5/27.6 fail before / pass after under the EPERM shim
    Full verdict

    CANNOT-TEST (needs a file whose group cannot be re-set by the owner, e.g. inherited from a setgid dir of a foreign group on NFS/idmapped storage; on local Linux fs a no-op chgrp to the current gid is always allowed); tests unchanged on main, 9.0, 8.6

    • Platform unix
    • Tested main on Linux: [file attributes f -group <current>] succeeds; user namespaces are blocked here so the foreign-group case could not be set up
    • Cause tests/fCmd.test fCmd-27.5/27.6 re-set the first attributes returned, which on Unix are -group (and -owner); SetGroupAttribute does chown(path, -1, gid), which can fail with EPERM when the group name resolves to a gid the user may not set.
    • Prototype prototypes/17a26ddf66.patch (6 changed lines, tests fCmd-27.5/27.6 now drop -group with [dict remove]; fCmd.test passes)
    • Suggested ticket comment The tests pick -group, which is not always settable by the file owner. Proposed fix: drop -group from the attribute list in fCmd-27.5/27.6 (patch available, not committed).
    ⬜d19aef14872014Does not cross-compile for iOS 6 - patch included.CANNOT-TEST (iOS SDK) — first half FIXED since 8.6.2 (a41a431706, 2014-07-31)

    Open ticket d19aef1487 · 52. Portability Support · filed 2014-07-29

    Full verdict

    CANNOT-TEST (iOS SDK) — first half FIXED since 8.6.2 (a41a431706, 2014-07-31)

    • Platform build (iOS cross-compile)
    • Tested main built on Linux out of tree with --disable-load: compiles cleanly, load gives "dynamic loading is not currently available". tclLoadNone.c TclpLoadMemory now matches the tclInt.h prototype on main, 9.0 and 8.6.
    • Cause the TclpLoadMemory signature mismatch in tclLoadNone.c (the generic part) was fixed by jan.nijtmans in a41a431706 "Fix part of d19aef1487". The remaining part, unix/tclUnixPort.h #if defined(__APPLE__) && defined(__DYNAMIC__) → <crt_externs.h>/_NSGetEnviron, is unchanged on main/9.0/8.6. The reporter's TARGET_OS_MAC guard would not help anyway: TARGET_OS_MAC is 1 on iOS too, so the test would have to be !TARGET_OS_IPHONE / TARGET_OS_OSX. The ios branch (ba55cebfc9, 2014) was never merged and was closed in 2022.
    • Prototype none (cannot verify without an iOS SDK; iOS is not a supported platform)
    • Suggested ticket comment The --disable-load part was fixed in a41a431706. The remaining environ/_NSGetEnviron guard in unix/tclUnixPort.h is iOS-only and nobody can test it. Suggest closing as "Out of date" unless someone with an iOS toolchain can confirm that a !TARGET_OS_IPHONE guard is still needed.
    ⬜c4e230f29b2013unixeventfork-1.1 hangs on OS XCANNOT-TEST (macOS)

    Open ticket c4e230f29b · 01. Notifier · filed 2013-08-03

    Full verdict

    CANNOT-TEST (macOS)

    • Platform macOS
    • Tested unixforkevent-1.1 with -constraints nonPortable passes on Linux main. On macOS the test is still constrained {testfork nonPortable}, so it is skipped by default and no longer hangs the suite.
    • Cause (macOS) CoreFoundation is not fork-safe without exec. macosx/tclMacOSXNotify.c:AtForkChild now resets the locks, drops the inherited runLoop/source/timer and restarts the notifier thread, which may have improved this since 2013; not verifiable here. Branch bug-c4e230f29b (2013) only holds the test hardening, no notifier fix.
    • Prototype none
    • Suggested ticket comment The test is marked nonPortable and passes on Linux; macosx/tclMacOSXNotify.c AtForkChild has since been reworked to reset the run loop in the child. Someone with macOS could run make test TESTFLAGS="-file unixForkEvent.test -constraints nonPortable"; if it passes, this can be closed.
    ⬜3077617fff2010error trying to make on aix 5.3patchCANNOT-TEST (AIX); the cause is still present on main/9.0/8.6 by code reading

    Open ticket 3077617fff · 53. Configuration and Build Tools · filed 2010-09-28

    🧪 Prototype check (2026-10-24)

    3077617fff.patch: VERIFIED (configure level only; the AIX link itself is still untested)

    • Base main (tip d6dd6d0f71)
    • Before no AIX box. Simulated the AIX branch on Linux: tcl_cv_sys_version=AIX-5 unix/configure (gcc) -> Makefile LDFLAGS = $(LDFLAGS_OPTIMIZE), no -brtl (the xlc branch has it, the gcc branch does not)
    • Patch applied cleanly (1 hand-edited line in tcl.m4); ~/tcltk/autoconf-2.73/bin/autoconf -f regenerates a configure byte-identical to the patched one
    • After the same simulated configure -> LDFLAGS = $(LDFLAGS_OPTIMIZE) -Wl,-brtl; a plain Linux configure is unaffected (no brtl). Linking on AIX is not tested
    • Test none possible (build system; ran the simulated configure instead)
    Full verdict

    CANNOT-TEST (AIX); the cause is still present on main/9.0/8.6 by code reading

    • Platform build (AIX + gcc)
    • Tested read unix/tcl.m4 AIX-* (non-ia64) on main. For gcc, SHLIB_LD is ${CC} -shared -Wl,-bexpall with no -brtl. LDFLAGS="$LDFLAGS -brtl" is added only in the non-gcc (xlc) branch.
    • Cause without run-time linking (-brtl) the AIX linker resolves -ltcl8.5 only to libtcl8.5.a, never .so. So gcc ... -ltcl8.5 fails although libtcl8.5.so was built (reporter's ls confirms). The xlc path works because it adds -brtl.
    • Prototype prototypes/3077617fff.patch (1 hand-edited line in tcl.m4 + regenerated configure with autoconf 2.73: adds LDFLAGS="$LDFLAGS -Wl,-brtl" in the gcc branch). Untested; needs an AIX box.
    • Suggested ticket comment With gcc, tcl.m4 does not pass -brtl to the AIX linker (only the xlc branch does), so -ltcl8.5 cannot find libtcl8.5.so. Proposed fix: add -Wl,-brtl to LDFLAGS for gcc too. Needs testing by someone with AIX.
    ⬜3075025fff20108.6 head testsuite failures on windowsCANNOT-TEST (WinXP/MSVC6 test failures from 2010; partly fixed: filesystem-1.3/1.4 fixed by a47641a031, 8.6.6, per apnadkarni)

    Open ticket 3075025fff · None · filed 2010-09-24

    Full verdict

    CANNOT-TEST (WinXP/MSVC6 test failures from 2010; partly fixed: filesystem-1.3/1.4 fixed by a47641a031, 8.6.6, per apnadkarni)

    • Platform Windows
    • Tested none (Windows; the report is about 2010 trunk on WinXP)
    • Cause environment-dependent test failures (UAC/permissions, timing in async-4.3).
    • Prototype none
    • Suggested ticket comment filesystem-1.3/1.4 were fixed in 8.6.6 (a47641a031); the remaining ones (async-4.3, tcltest-2.0, winFCmd-1.38) were reported on WinXP/MSVC6 in 2010. Current failures on Windows 10/11 should get new tickets; suggest closing this one as out of date.
    ⬜2782814fff2009sample CVS snapshots contain old dataCANNOT-TEST (the sampleextension repository and the old CVS nightly tarballs are not available here). This is also not a Tcl core issue.

    Open ticket 2782814fff · 70. Sample Extension · filed 2009-04-28

    Full verdict

    CANNOT-TEST (the sampleextension repository and the old CVS nightly tarballs are not available here). This is also not a Tcl core issue.

    • Platform site (sampleextension snapshots)
    • Tested nothing to test locally. In 2014 aku checked sample-20101214.tar.gz and the sampleextension Fossil repo and found no nested stale "sample" directory. The CVS nightly snapshots no longer exist.
    • Cause n/a
    • Prototype none
    • Suggested ticket comment The CVS snapshots are gone, and the sampleextension Fossil repository has no stale nested copy (checked by aku in 2014). Closing as out of date. Please file sampleextension issues in its own repository.
    ⬜2093902fff2008Expect-spawn plink works in 8.4.16 but fails in 8.5.4CANNOT-TEST (Expect for Windows 5.43 binary from 8.4 copied into ActiveState 8.5.4; third-party package, error only in an attached screenshot)

    Open ticket 2093902fff · None · filed 2008-09-04

    Full verdict

    CANNOT-TEST (Expect for Windows 5.43 binary from 8.4 copied into ActiveState 8.5.4; third-party package, error only in an attached screenshot)

    • Platform Windows
    • Tested none
    • Cause likely an Expect binary built for 8.4 used under 8.5 (not a Tcl core issue); the error text is only in the screenshot attachment.
    • Prototype none
    • Suggested ticket comment This is about the Windows port of Expect (a separate package) copied from an 8.4 installation into 8.5.4; it is not a Tcl core bug. Closing as third-party.
    ⬜2032862fff2008headers broken for Mac OS X frameworkCANNOT-TEST (macOS); the cause is still present in main by code reading

    Open ticket 2032862fff · 50. Embedding Support · filed 2008-07-30

    Full verdict

    CANNOT-TEST (macOS); the cause is still present in main by code reading

    • Platform macOS
    • Tested code reading on main: private headers still use quoted includes of public headers (generic/tclPort.h #include "tcl.h", tclInt.h "tclPort.h"), and for the framework build unix/configure.ac puts them in a separate PrivateHeaders directory (PRIVATE_INCLUDE_DIR="${libdir}/PrivateHeaders"; install-private-headers in unix/Makefile.in copies them unchanged).
    • Cause unix/Makefile.in install-private-headers + framework layout from configure.ac.
    • Prototype none (macOS only; needs a decision on the framework layout, which should be tested on a Mac)
    • Suggested ticket comment Still applies to 9.x by reading the code: PrivateHeaders/tclPort.h includes "tcl.h", which lives in Headers/. Needs someone with macOS to test a fix (framework-style includes or -I to the Headers dir in tclConfig.sh).
    ⬜2010028fff2008TCL 8.5.2 test failures in x86_64 InterixCANNOT-TEST (Interix)

    Open ticket 2010028fff · 38. Init - Library - Autoload · filed 2008-07-03

    Full verdict

    CANNOT-TEST (Interix)

    • Platform unix (Interix/SUA on Windows Server 2008)
    • Tested none. Interix/SUA was removed from Windows (deprecated in 8/2012, gone in 8.1/2012 R2), and the reporter's toolchain ran an "l2ll" source rewriter turning long into long long, which plausibly explains the integer-conversion failures. Attachments not checked.
    • Cause unknown; most likely platform/toolchain-specific (LLP64 compiler with an LP64 rewriting wrapper).
    • Prototype none (obsolete platform)
    • Suggested ticket comment Interix/SUA no longer exists in any supported Windows version and the report is against 8.5.2; closing as out of date. Please reopen with a current platform if similar failures show up.
    ⬜2003130fff2008failing fCmd.test tests on new fileserverpatchCANNOT-TEST (NFS server that returns EISDIR); reproduced with an LD_PRELOAD shim that maps rename()'s ENOTEMPTY/EEXIST to EISDIR: fCmd-6.16, 6.18, …

    Open ticket 2003130fff · 37. File System · filed 2008-06-26

    🧪 Prototype check (2026-10-24)

    2003130fff.patch: VERIFIED

    • Base main (tip d6dd6d0f71)
    • Before no EISDIR-returning NFS here; own LD_PRELOAD shim work-a1/v1/shim/eisdir.so (rename(): ENOTEMPTY/EEXIST -> EISDIR). LD_PRELOAD=eisdir.so tcltest fCmd.test -constraints notWine -> fCmd-6.16, 6.18, 9.9, 9.13 FAILED (error renaming "td2" to "td1/td2": is a directory), 273 passed/28 skipped/4 failed. (-constraints notWine is needed: on the tip notWine is still undefined in fCmd.test, so these tests are always skipped, 300724b80c.)
    • Patch applied cleanly (13 changed lines)
    • After same command -> 277 passed/28 skipped/0 failed; without the shim 277/28/0 too; plain fCmd.test 260/45/0, fileSystem.test 141/21/0, cmdAH.test 16970/30/1. The one failure, cmdAH-20.5 (atime touch), is a timing flake that the patch does not touch: cmdAH-20.* passes 3/3 with and 3/3 without the patch
    • Test none possible without the shim (EISDIR cannot be triggered on a normal file system); the shim run of the existing fCmd-6.16/6.18/9.9/9.13 fails before / passes after
    Full verdict

    CANNOT-TEST (NFS server that returns EISDIR); reproduced with an LD_PRELOAD shim that maps rename()'s ENOTEMPTY/EEXIST to EISDIR: fCmd-6.16, 6.18, 9.9, 9.13 fail exactly as reported

    • Platform unix
    • Tested main (wt build) with work-a5/shim/eisdir.so, using the fCmd.test from branch fCmd-test-constraints (ed6326c9e3). On main these 4 tests never run at all: notWine isn't defined in fCmd.test, so they are always skipped. That's 300724b80c, fixed only on the unmerged branch fCmd-test-constraints (0049270a8b, 2026-10-12). 8.6's fCmd.test has no notWine constraint.
    • Cause unix/tclUnixFCmd.c:DoRenameFile maps ENOTEMPTY to EEXIST but passes through the EISDIR that some NFS servers return for directory onto non-empty directory (POSIX allows EISDIR only when the source is not a directory).
    • Prototype prototypes/2003130fff.patch (13 changed lines, no new test: EISDIR can't be triggered without a shim; verified with the shim and the branch's fCmd.test, 0 failures; main fCmd.test passes)
    • Suggested ticket comment The EISDIR comes from rename() on some NFS servers. POSIX allows it only when the source is not a directory. Proposed fix: DoRenameFile maps EISDIR to EEXIST when the source is a directory (checked with an LD_PRELOAD shim, fCmd-6.16/6.18/9.9/9.13). Note that on main these tests are currently skipped because of the missing notWine definition 300724b80c.
    ⬜1945538fff2008Stalling async reads with package TLSCANNOT-TEST (needs the TLS extension, not installed here; the reporter's stall scripts live on the old SourceForge TLS tracker #1945535, not in thi…

    Open ticket 1945538fff · 26. Channel Transforms · filed 2008-04-18

    Full verdict

    CANNOT-TEST (needs the TLS extension, not installed here; the reporter's stall scripts live on the old SourceForge TLS tracker #1945535, not in this repository)

    • Platform generic (TLS extension)
    • Tested none; ticket history read
    • Cause Per the 2010 analysis, the stall happens when a script-level read leaves data buffered inside the TLS layer (OpenSSL / TLS's own buffer) while the base socket has nothing more to read, so no readable event fires again. Generic Tcl has no way to see data held privately by a transform; a transform must schedule its own timer/notification (as tclZlib.c and tclIORTrans.c do). The "false positive readable" half was documented and split off as 3436609 (closed 2013).
    • Prototype none
    • Suggested ticket comment The remaining stall is a matter for the TLS extension's watch/notification of its own buffered data (the core zlib and reflected transforms schedule a timer for this). Suggest retesting with current TclTLS (2.x) and moving the issue to its tracker if it still stalls; otherwise close.
    ⬜1483314fff2006QNX6 - make test problemsCANNOT-TEST (QNX6)

    Open ticket 1483314fff · 52. Portability Support · filed 2006-05-07

    Full verdict

    CANNOT-TEST (QNX6)

    • Platform unix (QNX 6.3.2)
    • Tested code reading. The main complaint (buggy libc strtod) no longer applies: since 8.5 Tcl parses numbers with its own generic/tclStrToD.c, and there are no strtod() calls in generic/ or unix/ in 8.6 or main; compat/strtod.c/fixstrtod no longer exist. The other failures (clock, fCmd) were 8.4.12 environment issues per the reporter.
    • Cause QNX 6 libc strtod (8.4 only).
    • Prototype none
    • Suggested ticket comment Tcl 8.5+ no longer uses the C library strtod, so the suggested fixstrtod workaround is obsolete. If someone still builds on QNX, please report make test results for 8.6.18 or 9.0.4 in a new ticket. Closing.
    ⬜1113452fff2005Tk_MainLoop processes WM_DEVICE_CHANGE infinitely often.CANNOT-TEST (needs Tcl embedded in a thread of Macromedia Director MX 2004; nothing to reproduce with plain tclsh/wish)

    Open ticket 1113452fff · 02. Event Loops · filed 2005-01-31

    Full verdict

    CANNOT-TEST (needs Tcl embedded in a thread of Macromedia Director MX 2004; nothing to reproduce with plain tclsh/wish)

    • Platform Windows
    • Tested code reading of win/tclWinNotify.c (main): Tcl_WaitForEvent still does PeekMessage/GetMessage/TranslateMessage/DispatchMessage for any thread message, which is correct; a message that comes back endlessly has to be re-posted by some window of the host application (as davygrvy suggested), not by Tcl. 0x0118 is WM_SYSTIMER (caret blink timer).
    • Cause probably the host application's window in that thread re-posting WM_DEVICECHANGE; no Tcl defect shown.
    • Prototype none (the reporter's patch, dropping WM_DEVICECHANGE/0x0118, would break legitimate handlers)
    • Suggested ticket comment No reproducer outside Macromedia Director, and Tcl's notifier only dispatches what the thread's queue delivers; the endless re-posting most likely comes from a host window in that thread. Closing as out of date; please reopen with a standalone reproducer.
    ⬜1021580fff2004TCL Compile Errors on Tru64 UNIX 4.0FCANNOT-TEST (Tru64 UNIX; obsolete)

    Open ticket 1021580fff · 06. Time Measurement · filed 2004-09-03

    Full verdict

    CANNOT-TEST (Tru64 UNIX; obsolete)

    • Platform build
    • Tested read the attachment (present in the repository) and unix/tclUnixTime.c in 8.6/main
    • Cause K&R definitions of TclpGmtime/TclpLocaltime with TclpTime_t tt did not match the prototypes (promotmatch). TclpTime_t is gone, and the functions now have ANSI definitions taking const time_t *, so the error cannot recur. Tru64 is long dead and its cc is not supported by Tcl 8.6/9.
    • Prototype none
    • Suggested ticket comment The code in question (TclpTime_t, K&R definitions) was removed long ago, and Tru64 is unsupported. Suggest closing as Out of Date.
    ⬜956587ffff2004unixFCmd fails on OSF1 (current cvs)CANNOT-TEST (OSF1/Tru64 is a dead platform with no VM)

    Open ticket 956587ffff · 37. File System · filed 2004-05-19

    Full verdict

    CANNOT-TEST (OSF1/Tru64 is a dead platform with no VM)

    • Platform unix (OSF1/Tru64)
    • Tested unixFCmd-16.1 passes on Linux main
    • Cause unknown. On OSF1 chown() to the file's current owner gave EPERM. The test now has the unix notRoot constraints.
    • Prototype none
    • Suggested ticket comment Tru64/OSF1 is no longer a supported or testable platform, and unixFCmd-16.1 passes elsewhere. Propose to close as Out of Date.
    ⬜931847ffff2004TclFSNormalizeToUniquePath - incomplete?CANNOT-TEST (needs nested non-native VFSes whose normalizePathProcs hand off to each other). By code reading the limitation is still there: TclFSNo…

    Open ticket 931847ffff · 37. File System · filed 2004-04-08

    Full verdict

    CANNOT-TEST (needs nested non-native VFSes whose normalizePathProcs hand off to each other). By code reading the limitation is still there: TclFSNormalizeToUniquePath makes a single pass (native first, then each other FS once) and callers ignore the returned startAt.

    • Platform generic (VFS)
    • Tested code reading of main generic/tclIOUtil.c
    • Cause generic/tclIOUtil.c:TclFSNormalizeToUniquePath does not loop the way vincentdarley sketched in 2004.
    • Prototype none (design change; needs a nested VFS to test)
    • Suggested ticket comment Still a single-pass normalization in 9.x; it only matters for nested non-native VFSes. Keep it as a design note together with 941872, or close as RFE.
    ⬜922766ffff2004compile errors on LynxOS 3.1.0CANNOT-TEST (LynxOS)

    Open ticket 922766ffff · 53. Configuration and Build Tools · filed 2004-03-24

    Full verdict

    CANNOT-TEST (LynxOS)

    • Platform unix (LynxOS 3.1.0 PowerPC)
    • Tested code reading only. The tclClock.c part is obsolete (FormatClock and the timezone manipulation were removed by the 8.5 clock rewrite). tclPosixStr.c in 8.6 and main still has unguarded case EBADMSG: and case ENOTSUP:, so a platform where EBADMSG == ENOTSUP would still fail with "duplicate case value". The -lbsd link need is not handled in tcl.m4 (no LynxOS case).
    • Cause generic/tclPosixStr.c Tcl_ErrnoId/Tcl_ErrnoMsg: duplicate errno values on LynxOS 3.1.
    • Prototype none (platform unavailable, 20+-year-old OS; the attached 8.4.6 patch is against code that no longer exists apart from the EBADMSG guard)
    • Suggested ticket comment LynxOS 3.1 is long obsolete and nobody can test it; the clock part no longer applies. Closing unless someone can still build on a current LynxOS release and report.
    ⬜853302ffff2003compiler warning on Tru64Unix 5.1ACANNOT-TEST (Tru64)

    Open ticket 853302ffff · 52. Portability Support · filed 2003-12-03

    Full verdict

    CANNOT-TEST (Tru64)

    • Platform unix (Tru64 5.1A, vendor cc)
    • Tested none; the attached log is warnings from 8.4.5 sources, which have been largely rewritten since.
    • Cause n/a (warning list, no specific bug)
    • Prototype none (no compiler/platform; Tru64 is EOL since 2012)
    • Suggested ticket comment The warnings are against 8.4.5 sources that no longer exist in this form, and Tru64 is long out of support. Closing; current warnings from a modern compiler are welcome as a new ticket.
    ⬜837257ffff2003EBUSY failures in 'make test'CANNOT-TEST (needs an NFS-mounted build dir; no NFS here and mounting needs root)

    Open ticket 837257ffff · 37. File System · filed 2003-11-06

    Full verdict

    CANNOT-TEST (needs an NFS-mounted build dir; no NFS here and mounting needs root)

    • Platform unix (NFS)
    • Tested the fCmd tests named in the ticket pass on local ext4 with main
    • Cause unix/tclUnixFCmd.c:DoRenameFile just reports what rename(2) returns. Linux NFS clients give EBUSY when the target directory is still referenced (silly-rename/dentry in use), so this is filesystem behaviour, not a Tcl bug. DoRenameFile maps only ENOTEMPTY/EIO/EINVAL.
    • Prototype none (mapping EBUSY to EEXIST would hide real "busy" errors)
    • Suggested ticket comment Tcl only reports the EBUSY that rename(2) gives over NFS. The tests pass on local filesystems. Propose to close unless someone can still reproduce it on a current NFS setup.
    ⬜823330ffff2003decomposed unicode handling not transparent in glob on OSXCANNOT-TEST (macOS)

    Open ticket 823330ffff · 44. UTF-8 Strings · filed 2003-10-14

    Full verdict

    CANNOT-TEST (macOS)

    • Platform macOS
    • Tested code/doc reading only. Tcl still does no normalization on file names. Tcl 9.1 adds [unicode tonfc/tonfd] (TIP 726, main only, in 9.1.0), the script-level workaround the 2003 discussion asked for. On APFS (macOS 10.13+) names keep the form they were created with; the fully decomposed names this ticket is about came from HFS+.
    • Cause HFS+ stores names as NFD; [glob] returns them unchanged, so [string equal] with NFC literals fails.
    • Prototype none (macOS-only; any fix is a design decision on normalizing native names)
    • Suggested ticket comment Since 9.1, [unicode tonfc] (TIP 726) lets scripts normalize names returned by [glob]; APFS no longer forces NFD. Suggest closing unless someone wants Tcl to normalize native names itself.
    ⬜811991ffff2003HP-UX 11i looks for man pages in wrong place.CANNOT-TEST (HP-UX)

    Open ticket 811991ffff · 53. Configuration and Build Tools · filed 2003-09-24

    Full verdict

    CANNOT-TEST (HP-UX)

    • Platform unix (HP-UX 11i)
    • Tested code reading: unix/Makefile.in still installs section n pages into $(mandir)/mann with .n suffix (MANN_INSTALL_DIR), no HP-UX special case.
    • Cause HP-UX man(1) uses mannew/.new for section n; Tcl installs mann/.n everywhere.
    • Prototype none (platform unavailable; HP-UX 11i is end of life)
    • Suggested ticket comment Still installs into mann/*.n on all platforms. HP-UX users can set MANN_INSTALL_DIR at install time and rename; given HP-UX's status we don't plan a special case. Closing unless someone with HP-UX submits a patch.
    ✅219158ffff2000Test fCmd-9.4 relies on buggy rename implementationCANNOT-TEST (OpenBSD) — fCmd-9.4 passes on Linux (main 2026-10-22); the failure needs a rename(2) that requires write permission on a renamed direc…

    Open ticket 219158ffff · 37. File System · filed 2000-10-26

    ✅ Since triage (updated 2026-11-11)

    Status now: Closed (Fixed)

    • 🔧 2026-10-29 check-in e373a1adc9 on fCmd-test-constraints, in main: Rebase fCmd-test-constraints to main. Fix [219158], 300724b80c and 852a17eeb9: fCmd.test constraints (read-only directory rename, the lost notWine definitio
    • 🔧 2026-11-10 check-in 7aa93727ce on fCmd-test-constraints, in main: Rebase Fix [219158], 300724b80c and 852a17eeb9: fCmd.test constraits
    • 🔧 2026-11-11 check-in f9dfe7796a on trunk, in main, 9.0: Fix [219158], 300724b80c and 852a17eeb9: fCmd.test constraits
    • 🔧 2026-11-11 check-in 8068a3d2b3 on core-9-0-branch, in 9.0: Fix [219158], 300724b80c and 852a17eeb9: fCmd.test constraits
    • 💬 2026-11-11 jan.nijtmans: Fixed in f9dfe7796a and core-9-0-branch. Had to add an additional notOSX constraint on fCmd-9.4.2, that's all.
    Full verdict

    CANNOT-TEST (OpenBSD) — fCmd-9.4 passes on Linux (main 2026-10-22); the failure needs a rename(2) that requires write permission on a renamed directory

    • Platform unix (OpenBSD; per comments also macOS 10.5)
    • Tested main (c092dae4af, 2026-10-22) and branch tip ed6326c9e3 on Linux: fCmd.test 308 total, 0 failed on the branch
    • Cause tests/fCmd.test fCmd-9.4 expects file rename of a 555 directory to succeed; POSIX allows EACCES (OpenBSD does this on purpose since 2013)
    • Existing work branch fCmd-test-constraints (tip ed6326c9e3, 2026-10-12), unmerged (not in trunk/9.0/8.6); fossil merge --dry-run into current trunk gives 1 conflict in tests/fCmd.test, so it needs a rebase
    • Prototype none (branch exists)
    • To run on OpenBSD make test TESTFLAGS="-file fCmd.test" on main (fCmd-9.4 should fail) and on fCmd-test-constraints (fCmd-9.4.2 runs with renameNeedsDirWrite, no failure)
    • Suggested ticket comment none needed until the branch is rebased; then: "Branch [/timeline?r=fCmd-test-constraints|fCmd-test-constraints] rebased onto trunk, fCmd.test passes on Linux; please confirm on OpenBSD."