Open tcllib tickets: Not tested

Triage of all 509 open and Pending tickets of tcllib (bugs, patches and feature requests) on 2026-11-15. Each ticket was checked against tcllib trunk on Linux with Tcl 9.1, 9.0, 8.6 and 8.5, using local servers for the network modules and a Windows 11 VM for Windows-only tickets. Outdated, superseded, duplicate and misfiled tickets are listed separately. Rows marked patch link to a prototype fix with a regression test (147 in all, in tcllib-triage/prototypes). Updated 2026-11-16.

Needs Windows 1

⬜1537469fff2006"make install" fails on cygwinNEEDS-WINDOWS (Cygwin). Probably still present, judging from the code.

Open ticket 1537469fff · Bug · Installation · filed 2006-08-09

Full verdict

NEEDS-WINDOWS (Cygwin). Probably still present, judging from the code.

  • Module build system (configure.ac, Makefile.in, installer.tcl)
  • Tested code read only. configure.ac sets CYGPATH="cygpath -w" for every *CYGWIN_NT* host, regardless of whether the tclsh is a Windows-native build or Cygwin's own (Unix-platform) Tcl. The 2014 log shows the result: Cygwin's tclsh8.5 gets C:\Users\...\lib\tcllib1.16, treats it as a single relative path component, and file mkdir fails with "no such file or directory".
  • Cause configure.ac:4-11 chooses CYGPATH by uname -s instead of by the Tcl platform. With Cygwin's Unix-built Tcl, CYGPATH should be echo.
  • Prototype none. The fix belongs in configure.ac (plus regenerated configure) and needs Cygwin to verify. A regression test is impossible in the Tcl testsuite.
  • Suggested action keep open. Test on Cygwin with ./configure && make install. Fix by making CYGPATH echo when $TCLSH_PROG reports tcl_platform(platform) eq "unix".
  • Suggested ticket comment The 2014 log shows Windows paths (cygpath -w) handed to Cygwin's Unix-built tclsh, which cannot use them. configure.ac still picks cygpath -w for every Cygwin host. It should depend on the tclsh's platform instead.

Cannot test 4

⏳e599462c222019Update smtp client libraryCANNOT-TEST (no tls package here, external servers not allowed). From the code: hunk 1 (an explicit tls::handshake after tls::import for -tlsimport…

Open ticket e599462c22 · Bug · mime · filed 2019-10-05 · Pending (resolved, waiting for confirmation)

⏳ Since triage (updated 2026-11-16)

Status now: Pending

    Full verdict

    CANNOT-TEST (no tls package here, external servers not allowed). From the code: hunk 1 (an explicit tls::handshake after tls::import for -tlsimport) is not applied, and may matter, since the socket is made non-blocking with no handshake while the server waits for a ClientHello. Hunk 2 (adding the first EHLO line to args) is wrong: that line is the greeting, not a keyword.

    • Module smtp 1.5.2 (mime/smtp.tcl)
    • Tested code review; aku's 2019 logs (in a6f697a7e8) showed the same result before and after the patch
    • Cause smtp.tcl ::smtp::initialize_ehlo socket open path. Note: the handshake there runs on a socket -async channel that may not be connected yet.
    • Prototype none (needs a TLS server and the tls package to verify)
    • Suggested action keep open for someone with tls to test hunk 1 only. Related: a12bff914d (direct TLS patch).
    • Suggested ticket comment The second hunk is not needed: the first EHLO line is the greeting, not an extension. The explicit handshake for -tlsimport may still be useful, but it needs testing against an SMTPS server; see also a12bff914d.
    ⬜7554bfed302018IPv6 nameservers are not supported in UDP modeCANNOT-TEST (tcludp and ceptcl are not installed here). By code, still broken on trunk: udp_open is called without ipv6. rkeene's branch fix-dn…

    Open ticket 7554bfed30 · Bug · dns · filed 2018-12-10

    Full verdict

    CANNOT-TEST (tcludp and ceptcl are not installed here). By code, still broken on trunk: udp_open is called without ipv6. rkeene's branch fix-dns-udp-ipv6-7554bfed30 (31f9e28d7c, 2018-12-31) uses udp_open ipv6 for nameservers containing :. It looks right, but it was never merged and is untested.

    • Module dns 1.6.1
    • Tested code inspection of trunk and the branch
    • Prototype none (the branch exists; a test needs tcludp)
    • Suggested action test the branch with tcludp, then merge
    • Suggested ticket comment Still unfixed on trunk. The experimental fix on branch fix-dns-udp-ipv6-7554bfed30 looks correct and needs testing with tcludp before merging.
    ⬜e9886b57282016udpcluster not passing testsCANNOT-TEST (tcludp is not installed for any of our Tcl builds, so testsNeed udp skips the file; the tests also need working LAN broadcast/multic…

    Open ticket e9886b5728 · Bug · odie · filed 2016-10-06

    Full verdict

    CANNOT-TEST (tcludp is not installed for any of our Tcl builds, so testsNeed udp skips the file; the tests also need working LAN broadcast/multicast between 3 processes)

    • Module udpcluster 0.3.4 (trunk; the 0.4 rewrite of 2019 is only on the hypnotoad/sobyk branches)
    • Tested sak test run modules/udpcluster on 9.1 and 8.6 → 0 tests (skipped, no udp package).
    • Cause per aku (2017) only on 8.6.4: cluster::ping fails in the test setup ("Could not locate nns@<mac> on the network"), so it depends on the environment's network.
    • Prototype none (needs tcludp and a network)
    • Suggested action keep open until someone with tcludp runs the tests; close as outdated if they pass on 8.6.x/9.x.
    • Suggested ticket comment Only seen with 8.6.4 in 2017, and it depends on UDP broadcast on the local network. Can someone with tcludp re-run the udpcluster tests on a current 8.6 or 9.0?
    ⬜1224755fff2005ftp mode active/passive performanceCANNOT-TEST (firewall/DNS environment; no delay on loopback)

    Open ticket 1224755fff · Bug · ftp · filed 2005-06-21

    Full verdict

    CANNOT-TEST (firewall/DNS environment; no delay on loopback)

    • Module ftp 2.4.14
    • Tested active and passive NList against ftpd, about 45 ms each.
    • Cause unknown. Possible contributors: firewall handling of the server's connect-back in active mode, and OpenActiveConn calling fconfigure -sockname on every active transfer, which does a reverse DNS lookup of the local address.
    • Prototype none
    • Suggested action close as Works for me (environment-specific); optionally cache LocalAddr per connection
    • Suggested ticket comment No delay is reproducible locally. A 40-second stall only in active mode usually means the firewall blocks or delays the server's connect-back, or reverse DNS of the local address is slow. Passive mode is the right choice behind firewalls.