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 Outdated, superseded, duplicate and misfiled tickets are listed separately.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.
RFE-OPEN (feasible: a flat alphabetical list of packages/manpages would be generated in support/devel/sak/doc/doc_auto.tcl)
RFE-OPEN (large: PRECIS (RFC 8264) is a different framework, with IdentifierClass/FreeformClass, derived properties from Unicode data and profiles like RFC 8265/8266)
RFE-OPEN (implemented on branch e15c2a7c-json-write-hao, tip d6d4f017f5, not merged)
json-write-6.2. The last one (string-null default) should be 6.4.
string-null maps the empty string to null, so code that uses it cannot write an empty JSON string. This is documented, but it is a design choice the maintainer should accept explicitly.
-css. For now, #' #include "file.md" can serve as a per-file header/footer workaround. The report…RFE-OPEN (feasible and small). mkdoc only has -css. For now, #' #include "file.md" can serve as a per-file header/footer workaround. The reporter is the mkdoc author, and oehhar invited him to contribute.
RFE-OPEN (trivial. Maintainer decision: drop the dead block, or deprecate/remove the three undocumented helpers)
json::dict2json, json::list2json and json::string2json have no callers and no docs.
dict2json. Its if {0 && ...} quoting block has been dead since 2006 (0de9029772). The helpers also produce invalid JSON for keys or strings that contain " or \, because string2json does no escaping.
if and say in a comment that json::write is the supported writer.
if {0 && ...} block now and deprecating the helpers in favour of json::write.
{*}$progress instead of eval [concat ...], and needs docs and a test)RFE-OPEN (feasible, code supplied; should use {*}$progress instead of eval [concat ...], and needs docs and a test)
create has no progress option on trunk.
{*}$cmd total done), and add/untar may want the same option.
justify still accepts only left/right/center)RFE-OPEN (patch not applied; justify still accepts only left/right/center)
RFE-OPEN (new nonterminal modes; overlaps with eb8c1e69ba "prune")
RFE-OPEN (design proposal: extra per-nonterminal data and grammar-embedded AST actions; feasible, but it is a design task)
RFE-OPEN (docs request; the suggested text is largely in the ticket and could become a new pt "grammar design tips" page)
-directory dir option for create/add that resolves files relative to dir and stores the relative names, like tar -C;…RFE-OPEN (feasible: e.g. a -directory dir option for create/add that resolves files relative to dir and stores the relative names, like tar -C; about 15 lines plus docs)
create accepts only -dereference/-chan/-gzip. add -prefix prepends a string but cannot strip one.
-directory option (or separate on-disk path and archive name) for create/add would cover it.
RFE-OPEN (feasible and small). The second variant, which joins the group automatically when the nameserver is a multicast address, avoids hard-coded values. It needs tcludp and package require ip, but the dns module already uses ip. A full mDNS client would also need to collect several answers per query, which the current single-reply design does not do.
ircs:// via tls::socket.)RFE-OPEN (TLS support. picoirc 0.14.1 in the same module already supports ircs:// via tls::socket.)
cmd-connect uses a plain socket $host $port with no hook to substitute a TLS socket.
socketcmd config option or similar, untestable without a TLS IRC server)
config socketcmd ::tls::socket).
RFE-OPEN (feasible, but needs the author to contribute and maintain it; no activity since 2022)
--. A single data argument ca…RFE-OPEN (Priority 1 is largely done: md4, md5, sha1, sha256, ripemd128/160, crc32, cksum, sum and crc16 all accept --. A single data argument cannot trigger -file/-filename. Priority 2, option consistency (-hex vs -format), is an API-wide RFE)
-- "-file /etc/passwd" (hashes the string) and with "-hex", "-filename" etc. as the only argument. md4/md5/ripemd/crc32 reject an option-like single data argument with "wrong # args" (no file access). sha1/sha256 hash it.
--
--. Data starting with "-" can't make them read a file. Making -hex/-format consistent across modules is still open as an RFE.
1 - p**n; Arjen (2022) wanted to look at the general gambler's-ruin problem instead; nothing done since)RFE-OPEN (trivial one-liner 1 - p**n; Arjen (2022) wanted to look at the general gambler's-ruin problem instead; nothing done since)
1-p**n and isn't a gambler's-ruin computation
RFE-OPEN (feasible, small, low value)
math::polynomials::roots for degree ≤ 2 (real and complex, numerically stable) would be cleaner. Otherwise reject.
truncate exists only in Tcl 9.0+. O…RFE-OPEN (feasible; branch virtchan-1985a34255 has aku's implementation from 2022-04-12, not merged). Refchan truncate exists only in Tcl 9.0+. On 8.6 tip, a handler that lists truncate is rejected with bad method "truncate". So the method must be added only on Tcl 9, or the packages must require 9; the branch instead bumps everything to require 8.6, which does not help.
chan truncate: works on 9.1, fails at chan create on 8.6.
package vsatisfies [info patchlevel] 9.
RFE-OPEN (decimal.test still has only 2 tests)
tcllib package that provides the release version and the list of packages. sak already knows the …RFE-OPEN (feasible: the installer could write a tcllib package that provides the release version and the list of packages. sak already knows the package list.)
tcllib package (only tcllibc and tcllib::testutils).
tcllib package reporting the release version and contained packages seems feasible from sak's package list.
RFE-OPEN (no TLS support in ftp on trunk; patch not reviewable locally)
RFE-OPEN (POSIX-style behaviour by design; GNU-style permutation would be a new opt-in)
getoptions with argv fileA -f v stops at fileA and leaves fileA -f v (9.1)
-- handling). Permuting by default would break scripts that pass option-like words after operands.
RFE-OPEN (same planned "market" package as e6b2685388, never created; the formula N/sum(1/x) is just the harmonic mean)
RFE-OPEN (Arjen planned a separate "market" package in 2019, never created)
nvlist errors on e.g. text/plain or …RFE-OPEN (feasible). Still true: nvlist only looks at the query string for GET and at the body for POST, and nvlist errors on e.g. text/plain or application/json ("Unknown Content-Type"). pooryorick implemented option 2 (post/poststring/querystring) on the pooryorick/mime branches, but they were never merged and the pyk branch was closed in 2024-11 (51bb6cedcd). Option 1 (a backward-compatible querystring parser next to multipart) is small.
ncgi::reset "a=b" text/plain; ncgi::nvlist gives the error "Unknown Content-Type: text/plain" on 9.1 and 8.6
RFE-OPEN (feasible: comm::future is a small snit type; TclOO is available in 8.6+, but comm still supports 8.5, so it needs either dropping 8.5 or keeping snit there)
package require snit for comm::future)
RFE-OPEN (still true: apart from the filter encoder, ldap.test only checks argument errors)
mockAccept/mockRead, test ldap-30.0) that answers a rootDSE search and an extended request. It is the first test that exercises real protocol traffic, and it works with plain sockets on 127.0.0.1 without devtools/coserv.
RFE-OPEN (feasible but large: needs a new PE operator or literal syntax, plus support in the PEG grammar, verify, the Tcl/C PARAM engines, all generators and the docs)
RFE-OPEN (no prune mode; modes are still value/leaf/void). Closely related to 4a80a49118 ("New modes for nonterminals"), which proposes a never-a-node mode. AK noted it can be done as a post-parse AST transform.
imap4::fetch does not support raw FETCH items such as BODY.PEEK[1]: any unknown word, including "BODY" its…RFE-OPEN (feasible but not trivial). imap4::fetch does not support raw FETCH items such as BODY.PEEK[1]: any unknown word, including "BODY" itself, is treated as a header field name. The request therefore becomes BODY[HEADER.FIELDS (BODY.PEEK[1] FROM)], and the echoed response has nested brackets, which imaptotcl_symbol cannot parse. Its regex [\w\.]+\[[^\[]+\] splits it into BODY HEADER.FIELDS S ....
imap4::imaptotcl on 9.1. (BODY[1] {46}) parses fine. (BODY[HEADER.FIELDS (BODY.PEEK[1] FROM)] {46}) gives garbage, as reported. CANNOT-TEST against a real server (no IMAP server available; would need a fake server on 127.0.0.1).
fetch (default branch of the item switch) and imaptotcl_symbol (regex)
BODY.PEEK[1] is sent as a header name inside HEADER.FIELDS and the echoed reply confuses the parser. Supporting BODY[section] items in fetch is an enhancement still to be done.
---
Side findings (no ticket in this batch):
tclsh9.1 httpd.test aborts: "Error in ncgi.tcl: can't find package fileutil", because ncgi is loaded before fileutil in the support block). It works only through sak.
::fileutil::magic::filetype is missing, so serving any file that goes through ::httpd::mime-type (for example .html) returns 500.
return -code $result -options $opts rethrows read errors with an empty message.
RFE-OPEN (feasible but needs new options; NEEDS-WINDOWS to test)
init calls GetWin32Proxy http only. ProxyServer may hold http=...;https=..., and GetWin32Proxy already parses it per protocol (the answer to aku's question). The https_proxy environment variable is ignored on Unix as well.
filter, and tunnel_connect for https, use the same one.
ProxyServer uses the form http=h:p;https=h:p, and GetWin32Proxy can already pick https. Supporting it needs a separate https proxy setting (also for https_proxy on Unix).
RFE-OPEN (low value: anonymous submitter, no follow-up since 2016, needs an external browser/WebDriver and no test infrastructure here)
-callback override in the generated proc; the workaround in the ticket still works)RFE-OPEN (feasible: e.g. a per-call -callback override in the generated proc; the workaround in the ticket still works)
%name% (and %name:default%) URL tokens still work but are only documented under ::rest::substitute; the create_interface not…RFE-OPEN / DOCS. %name% (and %name:default%) URL tokens still work but are only documented under ::rest::substitute; the create_interface note sits in a [comment] at the end of rest.man. rest::describe still does not list them.
::rest::substitute on the url; describe looks only at req_args/opt_args)
RFE-OPEN (small, compatible. Prototype ready)
${log}::logproc warn [list ::msg warn] fails on trunk with Invalid cmd '::msg warn' - does not exist (9.1, 8.6). Found a related bug: when a child service is created after a new-style logproc is set on the parent, logging from the child passes the internal args to the callback: myp: <-_logger::service par::ch {hi child}>.
logproc, one-argument case: info commands $cmd checks the whole prefix, and the generated proc splices the prefix with [list %s ...]. That generated proc also skips _setservicename, which the other two forms call, so the child's -_logger::service marker reaches the callback.
linsert $prefix end $text, and wraps the call in _setservicename/_restoreservice. Both new tests fail without the fix. The log module passes 373/373 on 9.1 and 8.6.
-_logger::service args leaking to a parent's callback when logging from a child service.
RFE-OPEN (stale). Branch tkt-785d2954d4-jsonc (last 2da77f541a, 2015-06-08) was never merged. Trunk C code since moved to bison 3 / Tcl 9 and still builds lists and string objects.
fossil diff --from trunk --to tkt-785d2954d4-jsonc and code read (no critcl here)
tcl::process status $pid (TIP 462) returns the CHILDSTATUS of exec & children; still no portable way on 8.6)RFE-OPEN (now feasible on Tcl 9: tcl::process status $pid (TIP 462) returns the CHILDSTATUS of exec & children; still no portable way on 8.6)
exec sh -c "exit 3" & then tcl::process status → {1 {child process exited abnormally} {CHILDSTATUS pid 3}}. processman.tcl unchanged: spawn uses exec ... &, events only calls the onexit script, no exit code is kept.
::processman::events / onexit keep no status; hypnotoad (2015) said this needs C-level plumbing. Tcl 8.7+/9 now has it.
tcl::process exists)
tcl::process status returns the exit status of exec & children, so onexit could pass it to the callback without C code. 8.6 still has no portable way.
RFE-OPEN (feasible: the package is small and pure Tcl)
RFE-OPEN (feasible: a keyword-index-style generator in sak doc / dtplite; nobody worked on it since 2015)
RFE-OPEN (still missing on trunk; draft branch exists)
OpenActiveConn builds PORT with string map {. ,} on the local address, so with an IPv6 control connection it sends e.g. PORT ::1,p1,p2. OpenPassiveConn only parses the IPv4 227 reply. There is no EPRT/EPSV on trunk. aku's draft is on branch ak-tkt-b41b6f76b0-ftp-eprt (ad53730d6b, eb27679053, 2014-12, ftp "2.5"), never merged. No local IPv6 FTP server was available to test the branch.
OpenActiveConn / OpenPassiveConn are IPv4-only.
RFE-OPEN (feasible, same situation as 1852685fff; no patch was supplied)
RFE-OPEN (feasible, behaviour change)
option -x -default bad -validatemethod V then T create t succeeds; V is not called for the default.
options() in RT.type.typemethod.create without calling the validator.
file/dir classes via fileutil)RFE-OPEN (feasible; the patch on 3584575fff implements file/dir classes via fileutil)
string is classes plus enumerations.
indented/aligned pretty-prints…RFE-OPEN (low value; obsolete in practice). Pretty-printing a Tcl dict is not a JSON-parser concern. json::write indented/aligned pretty-prints JSON output.
RFE-OPEN (no SOCKS support in tcllib; the 2012 patch against 1.14 was never reviewed. It needs review for Tcl 8.5+/9 and a test suite with a local proxy.)
RFE-OPEN (large: OpenPGP packet format, key parsing, web of trust; nothing has been done since 2011)
RFE-OPEN (feasible; overlaps 1427906fff)
-maxsecs exists, used for the end-of-DATA reply (default 120). Connect/greeting/command waits are fixed (300–3600 s, capped at 600 per read).
talk/hear timeouts
RFE-OPEN (feasible: a -timeout option via the existing cget/configure plus a fileevent/vwait getline, about 40 lines. Untestable without a mock NNTP server, and none exists in the module.)
gets. The _configure proc exists now (it did not in 2010), so a -timeout option has an obvious home. The attached nntp.tcl is a full copy of 0.2.2 with getline/timeout added, and is not applied.
RFE-OPEN (the -des-mode/null padding request is questionable; silent key padding is unsafe). DOCS part still valid: des.man says -key "takes a binary string of 8 bytes" and never says that a 24-byte key selects 3DES.
htmlparse::parseAttributes proc that returns a name/value dict and handles quoted, unquoted and bare attributes)RFE-OPEN (feasible: a small htmlparse::parseAttributes proc that returns a name/value dict and handles quoted, unquoted and bare attributes)
RFE-OPEN (needs an asynchronous API redesign, e.g. callbacks or coroutines)
hear still does vwait ${token}(readable).
RFE-OPEN (partly done: point 2 is done, the matrix uses 1/0 per graphops.man. Points 1, 3 and 4 (separate node list, direction/multiplicity, combiner command) are not implemented.)
RFE-OPEN (feasible but incompatible: comm.tcl creates ::comm::comm at package load, and existing code relies on it)
comm::comm destroy if the default channel is unwanted.
RFE-OPEN (feasible and small. Needs a decision on where AD helpers go)
binary scan "x2SIi*" uses signed S/I, so an identifier authority of 2^31 or more gives a wrong value. Since Tcl 8.5 it should use Su/Iu/iu* and drop the manual overflow fix-up.
Su Iu iu*, Tcl 8.5+) and needs a home (ldap or a separate AD helper package) plus tests.
RFE-OPEN (vague; nothing implemented)
RFE-OPEN (feasible, no design exists; filed by the maintainer as an idea in 2009)
RFE-OPEN (feasible; still single value)
_lappend, __arc_lappend and __node_lappend, and tree_tcl.tcl _lappend, all take exactly key value. So do the C implementations (graph/methods.c, tree/m.c).
RFE-OPEN (contributed rewrite, never applied. It only supports RFC 977 commands, so applying it as is would drop the X* extensions trunk has.)
gets.
C:\TEMP C:\TMP \TEMP \TMP after TMPDIR/TEMP/TMP.RFE-OPEN (feasible, small; NEEDS-WINDOWS to test). The Windows fallback list is still C:\TEMP C:\TMP \TEMP \TMP after TMPDIR/TEMP/TMP.
$env(LOCALAPPDATA)/Temp and $env(SystemRoot)/Temp, or derive the directory from Tcl 8.6+'s file tempfile, which uses GetTempPathW.
basedir method on trunk. The patch targets traverse 0.4 and has no tests. traverse cha…RFE-OPEN (patch not applied: no -notroot, -sort, -order or basedir method on trunk. The patch targets traverse 0.4 and has no tests. traverse changed a lot since then, with symlink/cycle fixes and callback fixes, so it no longer applies.)
Pad), so a -padding pkcs7 option would be about 30 lines and backward comp…RFE-OPEN (feasible). aes still pads only with NUL bytes (aes.tcl:509 Pad), so a -padding pkcs7 option would be about 30 lines and backward compatible. -hexkey is not really needed, because binary decode hex (8.6+) covers it; a doc example would be enough. The related des request is 3002617fff.
-key [binary decode hex $hexkey] (Tcl 8.6+).
RFE-OPEN (feasible: a new markup command in checker + all formatters)
const, term, cmd, example, ...).
RFE-OPEN (language extension)
alt/rep commands or display-syntax block exist. A workaround works: [opt "[arg option] [arg value] ..."] renders as ?option value ...?, and a double-quoted [vset NAME "[...markup...]"] stores formatted markup for reuse.
[vset name "markup"] (doctools since 2003) acts as a parameterless macro; there are no parameterized macros)RFE-OPEN (partly covered: [vset name "markup"] (doctools since 2003) acts as a parameterless macro; there are no parameterized macros)
[vset OPTS "[opt "[arg option] [arg value] ..."]"] then [vset OPTS] inside [call] renders correctly in text and html on 9.1. With braces, the markup stays literal.
RFE-OPEN (not implemented. Would need an API redesign. The 2008 maintainers suggested a major version)
foo::v1, foo::v2) gives most of this today.
deserialize serial ?node? that inserts under a node; needs node-name clash handling and both Tcl and critcl impleme…RFE-OPEN (feasible: a "graft"/deserialize serial ?node? that inserts under a node; needs node-name clash handling and both Tcl and critcl implementations, so moderate work)
deserialize takes only the serialization, and there is no graft method.
RFE-OPEN (the code is attached and the Tcl license fits, but it was never reviewed. It would need tests and docs, and overlaps fileutil::find/traverse.)
ip::version.)RFE-OPEN (feasible. The ip.tcl core already has IPv6 Normalize6/Mask6, so isOverlap could dispatch on ip::version.)
ip::isOverlap 2002:0a23:0100::/80 2002:0a23:0100::5 gives "invalid ip address"
RFE-OPEN (feasible; attached patch was never applied)
C (no =) inside a section is silently dropped by _loadfile, and lost on commit.
*=* lines; there is no state for "key without value".
solo array marks bare keys. They read as value "" (value with a default returns the default, as in the original patch). commit writes them back bare. set/delete/revert clear the mark. ini.man not updated yet.
value for bare keys (default vs ""), then apply the prototype with a doc note. Close 2037560fff as a duplicate.
RFE-OPEN (not implemented. It depends on knowing which commands a page defines, the same missing piece as 1798413.)
c_xref_keywords and c_get_title at manpage_begin)RFE-OPEN (feasible, small: fmt.html already has c_xref_keywords and c_get_title at manpage_begin)
"$title - $shortdesc" (moddesc). There is no <meta name="keywords"> or <meta name="description">.
<head> with only charset/viewport meta and the title.
RFE-OPEN (feasible but a design change of the dtplite per-module TOC layout)
name titledesc rows.
[call] entries get numeric anchors <a name="1">, <a name="2">, and the xref database knows nothing about c…RFE-OPEN (still reproduced on 9.1: [call] entries get numeric anchors <a name="1">, <a name="2">, and the xref database knows nothing about command definitions, so [cmd other] does not link to the page that defines it. This is a feature request, not a bug.)
#N links.)
RFE-OPEN (partly done: pdf/cdf/random-lognormal exist; income-distribution functions such as Lorenz/Gini were not added)
RFE-OPEN (still not supported. Related patch ticket 3495830fff, "First go at installing tcl modules", is also Open.)
RFE-OPEN (not implemented. It is easy to do as a counting hook in checker.tcl, but AK filed it as a self-reminder and no user asked for it.)
RFE-OPEN (partly done. The doctools2 rewrite of the index and TOC parts, from 2009-03, uses pluginmgr and snit for its import/export plugins. The main doctools (doc) format still uses the old safe-interp engine. doctools::doc from doctools2 was never written.)
RFE-OPEN (partly done. doctools2idx/doctools2toc are snit types. doctools.tcl is still the custom object system.)
RFE-OPEN (no such engine. It would be simple, about 50 lines based on fmt.null with fmt_plain_text collecting words.)
RFE-OPEN (no such engine. It is feasible in the same way as 1695126.)
export_docidx/export_doctoc for their formats, but there is no…RFE-OPEN (there is no doctools-to-doctools pretty printer. doctools2idx/toc have export_docidx/export_doctoc for their formats, but there is no equivalent for doc.)
RFE-OPEN (depends on 1695126/1695128, which do not exist. It is a list of ideas with no concrete request.)
RFE-OPEN (partly done. doctools2idx/toc parse into a canonical serialization, a tree/dict, and the exporters work on that. The doc format still uses the event-stream engine API.)
[sectref id] links to (sub)sections within a document, and dtplite links [term]s that …RFE-OPEN (partly done. Since doctools 1.3.4, 2008-05-15, [sectref id] links to (sub)sections within a document, and dtplite links [term]s that match another document's title. [sectref-external] only formats the text, with no link and no check. There are still no links to a specific section of another document.)
RFE-OPEN (feasible, ~25 lines + docs/tests). There is no asnGetTLV/asnTLV yet. It could be built on asnPeekTag (multi-byte tags) and asnGetLength, instead of the single-byte asnGetByte used in the proposal.
ecb cbc)RFE-OPEN (patch never applied; modes are still ecb cbc)
if ($cfb_mode), unbraced expr, and it only allows a short final block. It also predates the d56da1abca padding changes and would need rework, docs and test vectors.
RFE-OPEN (wiki page not checked, no network; no code was ever attached, and Tcl's own Unicode/msgcat support has since covered much of this)
ftp.tmp)RFE-OPEN (patch not applied; ftp.tcl has no STOU/STOT and inline Put without a name still uses ftp.tmp)
RFE-OPEN (feasible, small. No robots.txt support in tcllib.)
RFE-OPEN (hard: needs a protocol extension or reverse DNS; see aku's 2006 analysis); the manpage still documents the limitation
RFE-OPEN (feasible, about 15 lines)
initialize still uses a fixed hear $token 600 for the greeting. -maxsecs exists but only applies to the reply after the final ".". Other waits are fixed at 300/600/3600 s.
RFE-OPEN (not applied: RetrFast still splits the whole message for debug logging; there is no "tofile" retrieval mode)
file tempfile (8.6+).
file tempfile.
RFE-OPEN (feasible, low value)
[list $ip [file join $self libbench.tcl] ...]. A child interpreter cannot read a script inside a wrapped (starkit/zipfs) filesystem. Tcllib installs bench as a normal package directory, never as a .tm, so a stock install is not affected.
[lindex [file system $self] 0] ne "native", copy libbench.tcl to a temp file. A test would need a VFS mount.)
file system is not native.
RFE-OPEN (test-coverage task still valid)
can't read "children(y)": no such element in array (tree_tcl _walk).
walk currently fails with an internal "can't read children(...)" error in the Tcl implementation. The walk semantics under modification need defining before tests can be written.
RFE-OPEN (trivial to add)
info supports args/body/default/type/vars/options/methods/typevars/typemethods. There is no self.
$obj info if still wanted.
RFE-OPEN (there is still no table markup in checker.tcl. Feasible, but it needs a markup design and support in all 10 backends. Markdown/HTML/LaTeX/nroff(tbl) can do it, and text/wiki/tmml would need a preformatted fallback.)
_walk could take optional -from/-to keys and use the search path to start; about 20 lines plus docs)RFE-OPEN (feasible: _walk could take optional -from/-to keys and use the search path to start; about 20 lines plus docs)
walk cmd takes no range arguments (skiplist.tcl _walk).
walk always walks the whole list. A -from/-to key range would be easy to add on top of the existing search.
bibtex::format type key dict writer is about 30 lines)RFE-OPEN (feasible: a bibtex::format type key dict writer is about 30 lines)
Status now: Pending
RFE-OPEN (the 2018 comment refers to a9488befa9, which is on Poor Yorick's mime rewrite branch, never merged and closed by Jan in 2024. On trunk, getbody on a multipart still errors "MIME part isn't a leaf")
mime::getbody on multipart/form-data → error
RFE-OPEN (trivial to implement, but needs an API decision. hobbs asked for it to be optional)
printname still prints the "Total calls: 0" block for uncalled procs.
print ?-nonzero? ?pattern?)
sortFunctions calls or dump.
RFE-OPEN (feasible; the wiki implementation is small)
RFE-OPEN (feasible via the protocol extension mechanism, but the promised patch never came in 20 years; low value)
RFE-OPEN (code exists only in the old davkit starkit; it depends on TclXML/expat and a modified httpd)
RFE-OPEN (workaround works. Would be mostly covered by the 751707e0ae prototype)
${log}::logproc level argname body form with variable service still works on 9.1, and gives the child's name when a child service inherits it (Logging service bar::kid : fromkid). That form is documented as deprecated.
${log}::logproc error [list customized $service] pass the name as a prefix argument.
$service.
RFE-OPEN (pop3 is still fully blocking; feasible today with coroutines, but it means a large rewrite of every command, not a small patch)
RFE-OPEN (trivial to add, but low value: the submitter agreed that auto_execok does the same)
fileutil::which on trunk
quoteString would make ncgi depend on Tcl's http package, and the output differs anyway (http encodes space …RFE-OPEN (low value). Reusing http's quoteString would make ncgi depend on Tcl's http package, and the output differs anyway (http encodes space as %20, ncgi as +, and http leaves -._~ unescaped). The useful part of the request is matching http's UTF-8 behaviour, and prototype 1f29f7baf9 does that.
http::formatQuery a "ö -._~\n(" gives a=%C3%B6%20-._~%0D%0A%28 on 9.1 and 8.6
error branch writes a hardcoded plain-text block. A callback option or errmode, e.g. errorcmd, would be a small c…RFE-OPEN (easy: HandleError's error branch writes a hardcoded plain-text block. A callback option or errmode, e.g. errorcmd, would be a small change.)
RFE-OPEN (still no general XML parser in tcllib; tDOM is the de facto standard and widely packaged)
xml (TclXML) package, see closed ticket 28919a4f4e.
RFE-OPEN (feasible, but needs an API design: callbacks and a fileevent-driven state machine. Not a small patch.)
gets $sock in fetch/response), with no -command or progress callback.
RFE-OPEN (no code ever contributed; RSS has declined in importance)
RFE-OPEN (no code ever contributed; would need an XML parser, see 934702ffff)
RFE-OPEN (low value: FXP is disabled on most servers for security reasons). Same request as open RFE 1813193.
RFE-OPEN (feasible, small, but no interest in 20+ years; the same is easily written with control::do and while)
RFE-OPEN (low value; wiki page not checked, no network; would also need a licence check of wiki code)
list covers brace quoting, and the backslash map is a one-liner)RFE-OPEN (trivial, but low value: list covers brace quoting, and the backslash map is a one-liner)
list already produces properly quoted words, and no further helpers were proposed in 20 years. Closing.
RFE-OPEN (legal/policy question, needs a maintainer decision. It is not a code bug.)
RFE-OPEN (feasible but not small. It needs a socket-command option like pop3/smtp, AUTH TLS, PBSZ/PROT P, and TLS on the data connections too)
RFE-OPEN (author's resolution: use snit::macro / configuremethod)
RFE-OPEN (vague; low value)
::T2::Snit_methodm ... in snit 1, uplevel frames in snit 2).
RFE-OPEN (still relevant: there is a concrete clash now)
man/mann) with the man pages of Tcl 8.6/9.0/9.1 and the system. unicode.n (tcllib stringprep "unicode") clashes with Tcl 9.1's new unicode(n) command page. There were no other clashes with Tcl. Generic names (graph.n, queue.n, matrix.n, log.n, …) can still clash with other packages.
walk still visits every node at most once (graph_tcl.tcl visited array, graph/walk.c))RFE-OPEN (not implemented; walk still visits every node at most once (graph_tcl.tcl visited array, graph/walk.c))
info commands already covers it, and with TclOO info object isa object)RFE-OPEN (trivial, but info commands already covers it, and with TclOO info object isa object)
tree exists checks nodes, not trees.
[llength [info commands myTree]])
info commands myTree) answers this directly, and none of the other struct containers offer a class-level exists either. Closing.
comment, arg etc. still take exactly one argument, see checker.tcl proc comment {text}. Unresolved design question from AK's 2004 c…RFE-OPEN (comment, arg etc. still take exactly one argument, see checker.tcl proc comment {text}. Unresolved design question from AK's 2004 comment: is [arg a b c] one argument or three?)
manpage_begin title section version still requires all three arguments, checker.tcl:274. There is no alternative document-start command.)RFE-OPEN (manpage_begin title section version still requires all three arguments, checker.tcl:274. There is no alternative document-start command.)
section/version optional would be a small checker+backend change, but every backend uses them in headers.
RFE-OPEN (feasible, moderate: expand would need to read input from a stack that macros can push onto. Not implemented. The submitter is the maintainer, aku.)
RFE-OPEN (no docbook engine. Feasible with fmt.tmml/_xml.tcl as a template, but the requester himself suggested closing it in 2003.)
[require] lines must be in the header and are printed as one block before all [call]s. It would need checker …RFE-OPEN (still not possible: all [require] lines must be in the header and are printed as one block before all [call]s. It would need checker and backend changes to let require appear between calls.)
require foo 1.0, require foo 1.1 and two calls, in the text format on 9.1. The SYNOPSIS lists both requires first, then both calls.
require only in the header state. Backends collect requires and calls separately (c_held / synopsis).
RFE-OPEN (still not implemented; pop3d's TOP still builds the whole message and does split/lrange, pop3d.tcl ::pop3d::Transfer)