This is the mail archive of the mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: coreutils-5.3.0-1

Hash: SHA1

Attention: cygutils, gettext, and procps maintainers - see below for
packaging conflicts.

According to Corinna Vinschen on 1/31/2005 2:38 AM:
>>This release is marked as test for several reasons:
> Please consider to put this package out of "test" in a couple of days.

I forgot to mention one other reason - upstream designated 5.3.0 as
unstable, as compared to 5.2.1 being stable.  But it has worked pretty
reliably for me, so you are right that I should bump it to current after a
couple of weeks of no complaints on the cygwin list from early testers.

>>- pathchk regressed in behavior. In 5.2.1, `pathchk file/x' failed with a 
>>message that file is not a directory, but in 5.3.0 it silently succeeds.  I 
>>don't know if this is worth patching locally or waiting for cygwin 1.5.13.
> That's a Cygwin problem and will be fixed in 1.5.13, thanks to Pierre.

Then I won't bother writing a workaround patch for `pathchk' in the 1.5.12
timeframe unless someone asks for one.  Any idea when 1.5.13 comes out?
Will 1.5.13 support trailing spaces in filenames on a managed mount?

>>- I bundled coreutils kill as gkill, rather than omitting it altogether
> Oh boy.  Now we have *three* kills, kill from the Cygwin package, 
> prockill from the procps package and gkill from the coreutils package.
> Isn't that overkill?

Nice pun.  And remember that bash (but not ash) has a builtin kill,
different from all three executables.  About the only feature I can see
that GNU kill provides that the others don't is a nicer table listing with
`gkill -t', which is why I included it.

> Packaging looks good.  I've uploaded the package.  Please send an
> announce posting to cygwin-announce.  Don't forget to include the
> unsubscribe instructions in your announcement.

Hmm, I found some packaging issues (maybe we should fold a variation of
this check into the generic-build-script):

$ tar tjvf coreutils-5.3.0-1.tar.bz2 |
> sed -e 's,.*usr/,/usr/,' -e 's, -> .*,,' | grep -v '/$' |
> xargs cygcheck -fv | grep -v coreutils-5.2.1
/usr/bin/readlink.exe: found in package cygutils-1.2.5-1
/usr/lib/charset.alias: found in package gettext-0.14.1-1
/usr/lib/charset.alias: found in package libiconv-1.9.2-1
/usr/bin/uptime.exe: found in package procps-010801-2

I think /usr/lib/charset.alias should be owned by gettext.  How about
readlink and uptime, should they be switched to ownership by coreutils, or
should I rename them to greadlink and guptime?  It looks like I will have
to make a -2 release soon to get rid of these packaging conflicts.

- --
Life is short - so eat dessert first!

Eric Blake   
Version: GnuPG v1.4.0 (Cygwin)
Comment: Public key at
Comment: Using GnuPG with Thunderbird -


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]