[Pkg-zsh-devel] DZB: Debian Zsh Bugs - The big picture [really really long]

Frank Terbeck ft at bewatermyfriend.org
Thu Feb 10 20:20:52 UTC 2011


Hey there list, here's a list of all currently open bugs for debian's
zsh package.

I might make this a regular post (monthly or so) if people think it's
worthwhile. This list is quite long - brace yourself. If you want to
smaller, help closing bugs. Also, this list was written using emacs'
org-mode, which may help you reading it, too.

Actually, the text became so long, that I'd recommend org-mode. Load
the file, hit "Shift-Tab" twice, and you got a nicely outlined
overview. Move to any headline and hit tab repetively to toggle its
visibility.

This doesn't discriminate between serious, normal and wishlist bugs. I
only care about how hard it is to fix a bug.  Also, I truncated a few
bug titles (you'll see a "[...]" at the end of those lines).

Here it goes:


* Open Bugs for debian's zsh package

The bugs are sorted into fixed and outstanding bugs. Outstanding bugs
are split into easy, not-so-easy and hart. The bugs themselves in the
subcategories are sorted from old to new. Some bugs, while being open
in the BTS may in fact by fixed. Those are in the "Fixed in *"
categories.


** Fixed in current packages

If there are bugs listed in here, that means the bug has been dealt with
and for some reason wasn't closed by an upload (probably because of
missing Closes:).

*** #162291 zsh: Path completion fails for paths with ..

    Marked as fixed-upstream. And that's true in fact. Close this one
    therefore.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=162291

*** #273097 'unset nosuchvar' should leave $? as zero

    This is fixed upstream for a while.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=273097

*** #545916 /etc/zsh/zshrc:14: unknown parameter: terminfo

    I pretty sure this one is fixed upstream. It should at least be in
    version 4.3.11 of the shell (and should therefore be fixed in the
    current version of the package). I'll look up the actual fix later.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=545916

*** #570878 zsh: Please add completion for pbuilder/cowbuilder

    Clint's completion function is shipped with the shell. So this bug
    is fixed already.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=570878

*** #589035 please provide zsh completion for osc

    The completion is included upstream. This should be closed.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=589035


** Fixed upstream

Bugs in this category have been dealt with upstream. It may be wise to
backport the fix for the package if that is feasible.

*** #276187 zsh: Ineractive menu completion with tab doesnt work

    I can't reproduce this with a current git HEAD build. Anyone on
    4.3.11, who could check? It's trivial to do, really.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=276187


** Bug reports with uncertain status

These bugs (if any) may not even be bugs. They might be bugs for other
programs or simply the result of pebcak. In any case, we may not be
responsible and the status should be evaluated as soon as possible.

*** #240618 cd completion offers usernames when CDABLE_VARS is set

    Quoting the manual:

    CDABLE_VARS
      If the argument to a cd command (or an implied cd with the
      AUTO_CD option set) is not a directory, and does not begin with
      a slash, try to expand the expression as if it were preceded by
      a `~' (see the section `Filename Expansion').

    And "cd ~<tab>" would offer users. So this actually works as
    documented. Clint commented, that the problem is that the resulting
    command wouldn't work. But that's not true:

      zsh-~% setopt cdablevars
      zsh-~% cd ft
      ~ft
      zsh-~ft%

    So this is not a bug and should be closed.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=240618

*** #368936 portable shell xargs too hard to write [...]

    I'd like to close this one, wontfix. I don't see how this is a zsh
    bug at all. I got the feeling that the reporter is pissed that
    shell scripting can be a bitch at times. (Also, people should use
    "find -whatever -exec the-code +" - yes, plus. It's POSIX.)

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=607628

*** #517008 alias not expanded with zsh -c

    As far as I know, upstream is not going to change the current behaviour
    drastically. And since it seems that the POSIX standard allows for
    this to be handled the way it currently is, I don't think this is
    even a bug. Upstream will put a warning about the behaviour into
    its manual.

    I think this should be closed.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=517008

*** #526647 zsh won't accept vaild commands when used through ssh

    Clint doesn't believe this is a zsh bug at all. I agree.
    Also, the reporter didn't react at all.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=526647

*** #557025 mplayer tab completion order

    Not a bug. It's a matter of configuration. Clint provided the
    required snippet. This should be closed.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=557025

*** #559124 unnecessary files

    I'm sorry, I don't consider this a bug. And if anything it's a wishlist
    priority thing. To be frank, I consider it largely masturbatory.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=559124

*** #578466 Shouldn't correct to "rm" too easily

    The correction system is fairly hardcoded with little to no
    configurability. While I'm surprised `grem' was suggested to be
    corrected to `rm', to fix this for real would be quite a task.
    It may be easier to bring in a special case for "rm", but then
    we may soon be needing one for "del" as well. The reporter
    said he'd be okay if the bug were closed, since he didn't read
    zsh's prompts at all, which was the reason for the actual problem
    in the first place.

    I think we should close the bug.

    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=578466


** Outstanding bugs

Outstanding bugs are sorted into their own sub-categories, from easy
to really really hard.


*** Low-hanging fruit

These bugs should be fixable with low to reasonable effort.


**** #138691 completion for man should find filenames as well as [...]

     I vaguely remember this being fixed upstream. Needs to be checked
     and if it's indeed fixed, closed.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=138691

**** #138912 ssh hostname completion shoud use hostnames from [...]

     Ssh host name completion has been debated a couple of
     times. Different people like to do it differently. I don't think
     we should change the default behaviour here. IMHO this should be
     wontfix.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=138912

**** #161506 most manpages should move to section 7

     Personally, I don't think it should. The manpages are really done
     big manual split into many. If it were one fat zsh(1), nobody
     would complain. I think this is wontfix, too.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=161506

**** #253794 Make <(..) work for shell scripts

     I can't reproduce this. Albeit, I'm not using `mailstat'. I made
     a trival script:

      #!/bin/sh
      head -n8 "$1"

     Now doing this: sh t.sh <( zcat some_file.txt.gz )
     ...works.

     I don't know. Maybe the Linux kernel's behaviour with respect to
     /proc/self/... changed. In any case, I think this should be closed
     either as can't-reproduce or as wontfix. Comments?

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=253794

**** #257336 command correction should not suggest _ prefixed completion functions

     This was true and incredibly annoying. In recent versions this can
     be configured away:

     CORRECT_IGNORE='_*'

     We should suggest that to the user and close the bug.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=257336

**** #345427 zsh has a different default PATH from bash's

     Usually, I'd say "And how is that a zsh bug and not a bash
     bug?". But this is probably easy to fix if upstream could be
     persuaded. If there's no `PATH' variable, zsh uses a default of
     "/bin:/usr/bin:/usr/ucb:/usr/local/bin", which is less then ideal
     on a debian system. Even if upstream does not want to change, we
     could patch this up within the package. I suspect the needed
     change to be rather simple.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=345427

**** #346119 Non-ASCII letters are not accepted in variable names

     Didn't look at it very closely, but I'm pretty sure this is fixed
     upstream. Therefore, this should be easy.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=346119

**** #349582 /etc/skel/.zshrc remains after upgrade

     Okay, some debian developer will have to comment here. I don't know.
     Purely a packaging issue.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=349582

**** #374906 Should take care of combining accents

     Err... Okay. I think zsh does handle combining accents in recent
     versions. And it seems the reporter concurs. Then they start talking
     about displaying the character. but if zsh correctly outputs the
     correct multibyte sequence, its job is really done and the terminal
     needs to do the right thing.

     I think this was a bug, but now is a terminal and/or font issue.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=374906

**** #418199 segfault with exceedingly long path

     This one has a fixed-upstream tag, due to a patch by pws. I
     vaguely remember this being committed upstream, too. Wouldn't
     hurt to retest, though.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=418199

**** #430146 please add .ssh/config Hosts to the hosts zstyle completion

     Clint gave a patch. The reporter didn't reply. .ssh/config is
     read by _ssh these days. Clint's patch is not applied upstream I
     think. We should poke the reporter.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=430146

**** #469458 backspace not working anymore

     Well. Keyboard handling in unix terminals is a bitch.

     To make things worse, there are different modes in which a terminal
     can be in. You can ask terminfo for the correct value of a key
     sequence, but for that the terminal needs to be in application
     mode. And even then, terminfo can be wrong and/or the terminal
     may behave wrongly.

     The problem is that zsh's line editor (zle) does not put the
     terminal into application mode by default. That's why we're
     doing s/[/O/ business in our zshrc when asking terminfo for
     key sequences (oh and that replacement is wrong for some
     terminals).

     The correct[tm] thing to do is to use a zle-hook to put the terminal
     into application mode when it's accepting interactive input. But
     the problem with that is, that it may break a lot of users' existing
     custom key bindings. So we're fairly screwed.

     To fix this particular bug, we should just bind '^?' and '^h' to
     whatever backspace should do. Unless someone has a convincing argument
     that that's a bad idea.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=469458

**** #481168 debconf question to determine whether bindkey -v should [...]

     I think this should be closed wontfix. I know that some people
     get confused. But this is historical behaviour, that is well
     documented. Also, zsh doesn't do this alone. Ksh93 does the
     same. The underlying problem IMHO is, that people don't realise
     that the shell's line editor is quite powerful and that there is
     a vi-mode which could be automatically chosen if they use vi in
     the first place.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=481168

**** #489646 Wish for /etc/zsh/completion.d

     I agree with Clint, that people should submit completions
     upstream so everyone can benefit. That being said, it's trivial
     to add a directory to $fpath via the global zshenv.

     However, it's *NOT* going to be in /etc. That's just plain
     wrong. We've seen what that leads to with grml-etc-core in the
     grml project before. FWIW, I'd vote for
        /usr/share/zsh/Completion/Vendor/
     or something along that line.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=611175

**** #510358 please allow use as /bin/sh

     I'm against doing this. I wouldn't do this myself, and I'm a
     hardcore zsh-fan. We had this in grml for a while and it worked
     for a while until it didn't, because a package maintainer decided
     to call a function `repeat' and was unwilling to change for zsh
     (which was understandable, since it's a perfectly valid function
     name). In zsh it's also a reserved word. Recent versions have
     quite a number of issues like that fixed in sh-mode. But I'd
     still advise against it. Strongly. There are better shells to use
     as `/bin/sh'. Therefore, wontfix.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=510358

**** #525313 zsh: please add completion for latexmk

     Currently, latexmk uses the same completion as latex, etc. Namely
     _tex. I think that's enough to close this bug as being fixed.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=525313

**** #535851 completion does not work anymore with some [...]

     Apparently zstyle ':completion:*' file-sort access causes a problem
     for a user. The bug contains a patch by Clint. I can't tell if the
     user tested it or not. No reply, yet.

     If Clint's patch fixes the issue, this is trivial to fix. If not,
     it may become hairier.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=535851

**** #536459 Ctrl-ARROW gives ';5A' escape sequence instead of moving [...]

     Another key binding issue. Ctrl-Arrow may not even work on a lot of
     terminals. IMHO people expect a little much from a basic default
     setup.

     For details, why keys and unix terminals are a pain, see the comment
     for #469458.

     I'm mildly leaning towards marking this one wont-fix.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=536459

**** #538069 Completions for paq8l, par2, sipp

     The reporter sent in three new completions. Clint asked a
     question. The guy didn't respond. I didn't check if the
     completions are included upstream by now or not. If not, we need
     to contact the guy and see whether he feels like answering
     Clint's question.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=538069

**** #554880 global zshrc should source /etc/zsh_command_not_found if [...]

     I wholeheartedly agree with Clint here. This should absolutely
     stay in the control of the individual user. It's trivial to add
     and potentially tedious to get rid of again. Wontfix. Really.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=554880

**** #568333 bad %? value for prompt expansion

     The reporter says this didn't apply to zsh-beta back then. I
     suppose it's probably not in 4.3.11 anymore then. Didn't try
     yet. But this is easy to verify and close (unless the bug
     reappeared).

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=611175

**** #579209 insert-last-word problem after completion

     Contains a patch by pws. Don't remember if it's applied upstream.
     If it is, it's in the 4.3.11 release and this bug belongs in the
     "Fixed in current packages" category. Otherwise it's trivial to
     fix, since the fix is already there.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=611175

**** #582258 zsh-mime-setup disables usual completions

     Contains a workaround by Clint and a more elaborate by pws. I need to
     check if this was committed upstream. If it was this is done. If not
     it may be trivial to fix if pws's patch does the trick.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=582258

**** #600151 zsh terminal-type completion doesn't look in [...]

     This sounds easy, by a first assessment.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=600151

**** #605849 A backgrounded pipeline should return a status of zero

     Sounded hard when I first read it. The reporter says that zsh-beta
     is fine, which may mean that the fix is in 4.3.11. Need to check
     if that's true. If so, this is trivial.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=605849

**** #611175 VCS_INFO wrongly detects svn repositories

     Contains a patch by the reporter, which is correct (I'm vcs_info's
     upstream). Trivial to fix therefore.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=611175


*** Thouger nuts

Bugs in this category are probably a bit harder to handle or they may
require a conciderable amount of work/testing to be resolved.

**** #51640 [difficult] zsh: Position of the cursor to fix errors

     The oldest one there is. :)

     I actually like the idea quite a bit. And it shouldn't be
     impossible, albeit not easy. I may take a stab at that when I'm
     vacationing next month...

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=51640

**** #175467 zsh -n wrongly executes math expressions

     This still seems to be around. I don't think we can go forward
     without involving upstream.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=175467

**** #247556 zsh build gets stuck running tests

     I don't have that environment available to check. If it's still
     there, it's probably hard to fix.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=247556

**** #269769 sabcmd completion doesn't work

     Didn't check yet. But probably not easy. Otherwise the people
     involved would have figured it out by now.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=269769

**** #284191 would like autocompletion for autogen.sh [...]

     Clint said something along the lines of parsing autogen.sh and
     then disect a configure --help call from that.

     I don't think that'll work well. The autogen.sh scripts I've seen
     are the ones that actually generate configure so we can't call it
     yet. I admit, it's annoying that some of them call configure with
     default options, but that not really zsh's problem.

     If someone wants to take a try, go ahead.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=284191

**** #285266 zsh: ecasound completion doesn't complete file names

     Didn't look at it at length. But reporters that don't react are
     the best.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=285266

**** #315255 cvs commit completion breaks on spaces

     Need to test this one... If it's still there, it'll be hard.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=315255

**** #321494 [zsh/stat] problems with large files

     Didn't test this, yet.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=321494

**** #330883 For svn revert, _subversion completes deleted files incorrectly

     This spans over two years. No final judgement was made. If this is
     still a bug, it will require some work.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=330883

**** #342347 zsh: _cache_svn_status makes subversion completion fail [...]

     Another one that needs to be tested.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=#342347

**** #346162 zsh: jobs -p is not POSIX-compliant

     I think this should be a case for a fixed up jobs in sh-mode. In
     zsh-mode this will never be changed. The zsh behaviour predates
     the spec.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=346162

**** #351663 change to /etc/zsh/zlogin

     This blocks #344030. It's requested to include some code in the
     global `zlogin' file. The code is this:

    # This fixes the desktop-profiles corner-case where a graphical client is
    # started through an ssh -X session (in which the Xsession.d scripts aren't
    # run, so we need to make sure the profiles are activated according to the
    # specified settings at login).
    #
    DESKTOP_PROFILES_SCRIPT="/etc/X11/Xsession.d/20desktop-profiles_activateDesktopProfiles"
    if (test -f $DESKTOP_PROFILES_SCRIPT); then
      # testing SSH_CLIENT as the woody ssh doesn't set SSH_CONNECTION
      # also testing SSH_CONNECTION as the current ssh manpage no longer mentions
      # SSH_CLIENT, so it appears that variable is being phased out.
      if ( (test -n "${SSH_CLIENT}") || (test -n "${SSH_CONNECTION}") ) && \
         (test -n "${DISPLAY}"); then
        # zsh needs the shwordsplit option set otherwise activateDesktopProfiles
        # script wil error out
        if (setopt | grep shwordsplit); then
          source $DESKTOP_PROFILES_SCRIPT;
        else
          setopt shwordsplit;
          source $DESKTOP_PROFILES_SCRIPT;
          unsetopt shwordsplit;
        fi;
      fi;
    fi;

     Does anybody know what that's about? I don't even know what
     `desktop-files' is.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=351663

**** #374913 zsh: Completion should handle combining accents equivalents

     Probably very hard. Could be in "Nightmare", too.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=374913

**** #375959 zsh: env completion is broken

     I didn't check if Clint's patch from the discussion was applied or
     not. We should check if this still applies to current versions.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=375959

**** #381842 completion of "dpkg -s" no longer works after a Ctrl-C

     This has to do with caching completion results. Need to check if
     this is still an active bug. Clint's patches in the report didn't
     seem to work for the reporter.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=513511

**** #389111 autocompletion fails on non-printable characters in [...]

     Clint mentions where the bug is hidden most likely. Probably very
     hard to fix. The bug is still in my current git build.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=389111

**** #397334 zsh fails to figure out the current directory on bind mounts

     Fairly old; Seems to be still active. Not necessarily easy to fix.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=397334

**** #419233 error messages start with lowercase letters

     This one is still present. Quite cosmetic. We should ask upstream
     if they care. (I'll admit, it looks ugly when the error are in
     german).

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=419233

**** #438666 zsh/stat and symlinks: reverse the effect of the -L option

     We should ask upstream. There's a patch included, so this might
     be easier than it sounds.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=438666

**** #463507 Completion fails with: "_main_complete:161: [...]

     This seems to have morphed from one bug into another. The original bug
     was caused by a bug in the newuser module, which should be fixed. The
     new one apparently is still on.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=463507

**** #486283 Use description of modules from zshmodules

     I think the underlying idea makes sense. I also think Clint is
     right to want to solve this at build time. It may take some work
     to get working flawlessly.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=486283

**** #490099 zsh: autocompletion enforces order, apt-get remove [...]

     This one is still a bug in recent versions. Someone with
     reasonable compsys insight should be able to deal with it. Please
     don't look at me right now. :)

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=490099

**** #505301 Please add completion for TopGit using its shell [...]

     Most work for this bug needs to be done by topgit's upstream. I'm
     not putting this in "Low-hanging..." because it involves compsys
     and most people are afraid of it.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=505301

**** #510038 glob qualifer - doesn't work correctly on dangling symlinks

     We need to check if this one is still a bug in recent versions.

     Also, there's a double close mentioned, which we could report
     upstream.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=510038

**** #512045 "select" gives incomplete list

     No reply yet.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=512045

**** #513491 %a support in printf builtin

     This is still a bug in latest versions. Don't know how hard it
     would be to fix. I'd imagine the effort would be reasonable.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=513491

**** #513511 zargs: cannot fit single argument within size limit

     I've seen this myself from time to time, but I don't know how to
     trigger it reliably, yet. If someone catches it, a shell trace
     might be useful.

     This is marked as "important". I don't know why that was done.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=513511

**** #527171 segfaults on long environment variables

     This may be a problem with the host's `setenv()' call. We should
     take a look at this. If it's a zsh bug it may be reasonably easy
     to fix (pws looked at the involved code and suspected that setenv()
     is not signaling an error when it couldn't fit a value into the
     environment - so I suspect it may not be a zsh bug).

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=527171

**** #533527 Provided scripts fail when cshjunkiequotes is enabled

     I remember this one. The _vim case should be handled, IIRC. The others
     should probably be fixed too. The problem is that with this option set,
     zsh can't deal with quoted strings that span over a newline (no, I do
     not know why anybody would ever want to set cshjunkiequotes at all,
     but I guess that ship has sailed).

     I should dig up the conversation on upstream's mailing lists.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=533527

**** #537678 after putting a job in background, line edition does not work

     This one could be tough. Axel asked for moreinfo, so let's see what
     that leads us to. If the reporter doesn't react and we can't reproduce
     it, then this should be closed.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=537678

**** #570951 ssh completes to non-resolvable hostname from [...]

     If we were to fix this one as suggested, it would be quite a bit of
     work. The ssh completion is already a little bug-ridden, if you
     ask me. So this may not be trivial.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=570951

**** #590454 Completion doesn't work on symlink to . with ignore-[...]

     No reply yet. Doesn't sound impossible to fix.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=590454

**** #593426 Status of background jobs not updated

     From what I gather from the conversation, this sounds hard to get
     right. There's a patch from pws, which Bart Schaefer wasn't quite
     sure about. So... this one needs work.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=593426

**** #597062 supply host completion for sshfs

     _sshfs uses _user_at_host, AFAICS from a quick look. Rsync uses
     its own function for that. Don't know what's the best route to go
     here.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=597062

**** #603503 numerous infelicities in git completion

     Ah, the _git completion. Source fear, awe, shock and horror.

     There is a major update available, which should fix all the issues
     mentioned in this bug, AFAIK. The update is not committed upstream
     yet, because sf.net's CVS is still down.

     The update should be tested and checked for these issues. Which is
     why I'm sticking it into the "Tougher..." category. If you'd like
     to fix all problems (especially the performance related ones, that
     task could easily go into "Nightmare").

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=603503

**** #607615 Behavior of ":A" and ":a" isn't clear and seems wrong

     Didn't read the report entirely yet. Those are recent features and
     the author is on IRC frequently. We should see if this is actually
     a problem and if so, ask upstream for assistance. Probably not too
     hard.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=607615

**** #607628 prompt not immediately updated after window resize [...]

     This one sounds incredibly annoying to hunt. Depending on what
     sort of bug is behind this, it could be rather hard to get to. I
     had hoped this was another SIGWINCH bug (which was fixed), but
     when I looked at the submission date it became clear, that this
     is probably something else. Bummer.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=607628


*** Nightmare

If you don't know what nightmare means here, you never played Doom. Anyway,
the bugs in this category will probably take quite a bit of man power and
brains to handle. I doubt any of them can be resolved without involving
upstream.

**** #256895 [compctl] completion fails for programs with spaces [...]

     This is an upstream confirmed bug. I don't know if it still applies to
     newer versions of the shell. If it does, this is probably hard to fix.

     The workaround is to start using compsys already.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=256895

**** #288323 doesn't handle suspension of commands in conditional [...]

     If I'm reading Bart correctly, zsh does the most reasonable thing
     here. I didn't investigate any further. If we decide, this is a
     bug, it's probably incredibly hard to get right. I'd close it
     wontfix.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=607628

**** #289748 menu navigation is suboptimal

     It seems this can be worked around.

     Fixing is probably a little out of the question. You usually know
     if you should go after a problem or not be listening to pws's
     comments. And here he made a joke about a 900-line function that
     contains the heroic number of five comments. I won't touch that.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=289748

**** #300470 trap mechanism on command-line length limitation with [...]

     Bart is convinced this can be done in a preexec() hook. So that's
     probably true.

     Implementing the entire idea should be a shit-load of work.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=300470

**** #303623 CHECK_JOBS doesn't work when only one command was run since last Control-D

     This seems to have to do with terminals and timing. I'm not going
     near it. Braver people are welcome to try.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=303623

**** #353863 No response and huge CPU usage on some completion(s)

     Here's a rather old one from 2006. Compsys bugs usually suck to
     debug. I've seen zsh's completion hang occationally, too. But so
     far I've never been able to really but my finger on it.

     This report is there another time: #353870 with slightly more
     information from the reporter. This probably makes sense to
     forward upstream if we're able to reproduce.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=353863
     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=353870

**** #514857 zsh -c 'set -e; ! true; echo OK' fails

     pws says he's not touching this until the standard gets clarified. So,
     it's not clear whether this is indeed a bug or not. If it is, it's
     probably a bitch to fix.

     Is there a bug state that would be appropriate? Pending? *shrug*

     This one is merged with #519571 and #520101.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=514857
     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=519571
     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=520101

**** #527110 segfaults when term is one line tall while doing tab [...]

     I need to check if it's still reproducable. If so, it may be
     hard to fix.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=527110

**** #555957 failed to create hard link as lock file [...]

     This was forwarded already. I'm not sure if it still applies. If
     it does, this could be quite a lot of work.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=555957

**** #589300 multibyte character support broken again

     Damn, multibyte issues scare the crap out of me. We should forward
     this one if it's reproducable (which it probably is).

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=589300

**** #589339 zsh frozen with zombie child

     I've got no idea how to handle this at all. We could try getting in
     touch with the reporter and see what his thougts are.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=589339

**** #603816 git checkout completion performance regression

     Just when I typed in the comment for #603503, I'm stumbling across
     this one. Like I said, the performance issues are a major issue,
     with _git in large repositories. The source of the issues is not
     so much of a mystery, but fixing them is a *major* pain in the ass
     that will take a *loooooooooong* long time.

     Oh, don't be fooled into using the bash completion with git, though.
     While that seems to be possible with recent versions of that code,
     the completion is BY FAR more simplistic than what zsh's _git has
     to offer. I have to use it from time at work, and trust me, you'll
     be catching yourself yelling "Why the f*ck aren't you completing
     that for me!?".

     Anyway. This is a bitch to fix.

     http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=603816



More information about the Pkg-zsh-devel mailing list