[pacman-dev] [PATCH] makepkg: Place packages symlinks in build dir when DESTDIR is used

[Allan McRae]

Eric Bélanger wrote:
> On Wed, Nov 4, 2009 at 2:37 AM, Eric Bélanger <snowmaniscool at gmail.com> wrote:
>> On Wed, Nov 4, 2009 at 1:45 AM, Allan McRae <allan at archlinux.org> wrote:
>>> Eric Bélanger wrote:
>>>>
>>>> Here's the latest patch. I hope everything is correct.
>>>>
>>>> BTW, do you want me to continue sending patches as attachment too?  I
>>>> see that 'git send-email' doesn't so I guess you can grab the inline
>>>> patch or get it directly from git.  Is that correct?
>>>>
>>> The way I handle this is to just send them directly with git send-email
>>> using the "In reply to" value to keep the thread... um... threaded.  Any
>>> comments about the patch and what you have changed to fix things can go
>>> under the "---" after the sign-off.  Of course, you can inline them as you
>>> have and push them to a git repo somewhere and we can just pull from there.
>> OK. I'll do that. I was inlining them manually to keep everything in
>> the same thread.
>>
>>> Also, can you insert newlines into you commit message to keep them at about
>>> 80 characters per line.
>>>
>>>> Signed-off-by: Eric Bélanger <snowmaniscool at gmail.com>
>>>> ---
>>>>  scripts/makepkg.sh.in |   26 +++++++++++++++++++++-----
>>>>  1 files changed, 21 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/scripts/makepkg.sh.in b/scripts/makepkg.sh.in
>>>> index 92b0454..aaf576b 100644
>>>> --- a/scripts/makepkg.sh.in
>>>> +++ b/scripts/makepkg.sh.in
>>>> @@ -138,11 +138,17 @@ clean_up() {
>>>>                if [ -n "$pkgbase" ]; then
>>>>                        # Can't do this unless the BUILDSCRIPT has been
>>>> sourced.
>>>>                        rm -f
>>>> "${pkgbase}-${pkgver}-${pkgrel}-${CARCH}-build.log"*
>>>> +                       [[ "$PKGDEST" != "${startdir}" ]] \
>>>> +                           && rm -f
>>>> "${pkgbase}-${pkgver}-${pkgrel}-${CARCH}${PKGEXT}"
>>>>                        if [ "$PKGFUNC" -eq 1 ]; then
>>>>                                rm -f
>>>> "${pkgbase}-${pkgver}-${pkgrel}-${CARCH}-package.log"*
>>>> +                               [[ "$PKGDEST" != "${startdir}" ]] \
>>>> +                                   && rm -f
>>>> "${pkgbase}-${pkgver}-${pkgrel}-${CARCH}${PKGEXT}"
>>>>                        elif [ "$SPLITPKG" -eq 1 ]; then
>>>>                                for pkg in ${pkgname[@]}; do
>>>>                                        rm -f
>>>> "${pkg}-${pkgver}-${pkgrel}-${CARCH}-package.log"*
>>>> +                                       [[ "$PKGDEST" != "${startdir}" ]]
>>>> \
>>>> +                                           && rm -f
>>>> "${pkg}-${pkgver}-${pkgrel}-${CARCH}${PKGEXT}"
>>>>                                done
>>>>                        fi
>>>>                fi
>>> I am not happy with this way of cleaning up the symlinks to packages.  If
>>> the package is not a split package, then the package name will be
>>> $pkgname-...   and not $pkgbase (although, they are likely the same thing).
>>  I was following (maybe too closely) the log files removal as template
>> here. I'll change it to pkgname.
>>
>>>  I am also not sure what the second clean-up (in the "$PKGFUNC -eq 1" test)
>>> is doing that has not already by the one above it.
>> Yeah, it probably can be removed.
>>
>>> However, thinking about this more, we do not remove old packages when using
>>> --clean so why remove symlinks to old packages.  Only when package symlinks
>>> are pointing to packages that no longer exist in PKGDEST have we made a mess
>>> that needs cleaned up.  So how about something like:
>>>
>>> for pkg in ${pkgname[@]}; do
>>>  for file in ${pkg}-*-*-${CARCH}${PKGEXT}; do
>>>   if [[ -h $file & ! -e $file ]]; then
>>>     rm -f $file
>>>   fi
>>>  done
>>> fi
>>>
>> Sure. I'll use this instead.  I don't use the --clean option so I am
>> indifferent. I'll send a new patch tomorrow.
>>
>> Eric
>>
>>> Everything else in the patch is fine.
>>>
>>> Allan
>>>
>>>
>>>
> 
> Here's the new patch.  I struggled getting git to work so hopefully
> it's OK.  BTW, the 'git send-email' --in-reply-to option asks for an
> identifier. I can't seem to find it in the patches or email headers in
> gmail.
> 
> 
> 
> Signed-off-by: Eric Bélanger <snowmaniscool at gmail.com>
> ---

Pushed to my working branch.  I added a comment above the clean up of 
symlinks just to document what was happening.

Allan