-----Oorspronkelijk bericht----- Van: arch-dev-public-bounces@archlinux.org [mailto:arch-dev-public- bounces@archlinux.org] Namens Aaron Griffin Verzonden: vrijdag 21 december 2007 8:09 Aan: Public mailing list for ArchLinux development Onderwerp: [arch-dev-public] Fwd: [signoff] bash 3.2.0330-1 and filesystem 2007.11-4

Please retest. This bash package overrides the previous one. Patch level updated, profile.bash added to backup=() array.

Potential issue: Apparently /etc/profile is not replaced if modified. Can someone verify if this is fixed or not? I will investigate further as I fear it may be a pacman bug - if so, I'm guessing keeping etc/profile in BOTH backup arrays for the time being may fix this.

filesystem changes: Setup for FHS compliant man pages - FS#8839 Claim ownership of /etc/profile from bash - FS#4766 Remove LESSCHARSET env var from /etc/profile - FS#8877

bash changes: Update patch level to 033 Remove /etc/profile from package. Move bashisms to /etc/profile.bash - FS#4766

/etc/profile was also largely revamped. Please comment on the changes there if you'd like.

You still haven't looked at the issue with /usr/man/man* getting replaced by a symlink, which is not possible on a system where manpages are already installed.

On Fri, 2007-12-21 at 10:40 -0600, Aaron Griffin wrote:

On Dec 21, 2007 1:44 AM, Jan de Groot wrote:

...

You still haven't looked at the issue with /usr/man/man* getting replaced by a symlink, which is not possible on a system where manpages are already installed.

But that can be covered at a later date. The actual "support" for this is simply unsetting MANPATH. If a package installs man pages to /usr/man/man1 it will still be looked up just fine.

The makepkg 3.1 will no longer move man pages to /usr/man, so I don't think we should do any moving until that is out the door.

Could you explain to me how the symlink will help? It seems like, to me, we're just going to confuse the crap out of pacman when it tries to install/upgrade. With the current implementation, man pages will work regardless of where they are.

Hmm, I must have been mistaken. I read the diff from the commit list, not the whole PKGBUILD. There was something with these two lines coming in: mkdir -p usr/share/man/man{1,2,3,4,5,6,7,8} ln -s man3 $startdir/pkg/usr/share/man/man3x I mistakenly read the ln -s operation on the second line to be a symlink to /usr, where it's actually just a man3x symlink. One thing wrong with this package though: you built it with a broken makepkg, the manpage directories are all moved back to /usr/man :P

Potential issue: Apparently /etc/profile is not replaced if modified. Can someone verify if this is fixed or not? I will investigate further as I fear it may be a pacman bug - if so, I'm guessing keeping etc/profile in BOTH backup arrays for the time being may fix this.

I can verify that this wasn't fixed on my update. My /etc/profile was still my original one and I ended up extracting the etc/profile from the filesystem package by hand. Jason

Am Dienstag, 22. Januar 2008 04:27:29 schrieb Aaron Griffin:

On Dec 23, 2007 7:20 PM, Jason Chu wrote:

...

...

Potential issue: Apparently /etc/profile is not replaced if modified. Can someone verify if this is fixed or not? I will investigate further as I fear it may be a pacman bug - if so, I'm guessing keeping etc/profile in BOTH backup arrays for the time being may fix this.

I can verify that this wasn't fixed on my update. My /etc/profile was still my original one and I ended up extracting the etc/profile from the filesystem package by hand.

Bump now that pacman 3.1 is out the door?

Right, because using the old profile breaks man pages stored in /usr/share/man like the one which comes with pacman. -- archlinux.de

Am Mittwoch, 23. Januar 2008 15:22:00 schrieb Dan McGee:

Signoff, i686, both packages.

I think you have just signed-off a package which is not yet there. :-) (or don't we want to bump filesystem/bash?) -- http://www.archlinux.de

On Jan 23, 2008 10:39 AM, Pierre Schmitz wrote:

Am Mittwoch, 23. Januar 2008 15:22:00 schrieb Dan McGee:

...

Signoff, i686, both packages.

I think you have just signed-off a package which is not yet there. :-) (or don't we want to bump filesystem/bash?)

The ones in testing are fine, aren't they?