[arch-dev-public] [Patch] dbscripts: fix for any arch support
Hi, Here is a patch (only provided inline) to fix any arch support to the dbscripts. It will add any support for the sourceball script and probably fix the clean up scripts. It also set the BUILDSCRIPT as it will be no longer defined in makepkg.conf as of pacman 3.3. The patch for db-update is required as the any arch is treated as a special case (and db-update works with current /arch/config ). Eric --- /arch/config 2009-08-03 12:31:28.000000000 -0400 +++ config 2009-08-14 18:32:02.000000000 -0400 @@ -8,6 +8,5 @@ STAGING="$HOME/staging" TMPDIR="/srv/tmp" -ARCHES=(i686 x86_64) -# To be used if $BUILDSCRIPT is removed from makepkg.conf -#BUILDSCRIPT="PKGBUILD" +ARCHES=(i686 x86_64 any) +BUILDSCRIPT="PKGBUILD" --- /arch/db-update 2009-08-03 12:31:28.000000000 -0400 +++ db-update 2009-08-14 19:59:27.000000000 -0400 @@ -106,6 +106,7 @@ for current_arch in ${ARCHES[@]}; do + if [ "$current_arch" != "any" ]; then ftppath="$FTP_BASE/$reponame/os/$current_arch" ftppath_any="$FTP_BASE/$reponame/os/any" @@ -226,6 +227,7 @@ fi repo_unlock $reponame $current_arch + fi done if [ -n "$to_add_any" ]; then
On Fri, Aug 14, 2009 at 9:44 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
Hi,
Here is a patch (only provided inline) to fix any arch support to the dbscripts. It will add any support for the sourceball script and probably fix the clean up scripts. It also set the BUILDSCRIPT as it will be no longer defined in makepkg.conf as of pacman 3.3. The patch for db-update is required as the any arch is treated as a special case (and db-update works with current /arch/config ).
Eric
also, in /arch/misc-scripts/make-sourceball change: if /usr/bin/svn export -q "$SVN_PATH/$packagename" $packagename; then to: if /usr/bin/svn export -q "$SVNREPO/$packagename" $packagename; then For some reason, the variable changed name. I just noticed in git history that the any arch was added then removed from ARCHES, not sure why
On Fri, Aug 14, 2009 at 10:26 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
On Fri, Aug 14, 2009 at 9:44 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
Another change, will fix commit: make-sourceball: license skip is NOT a failure http://projects.archlinux.org/?p=dbscripts.git;a=commit;h=891e4f353d098a639c... This patch will remove skip licences from failed.txt --- /arch/misc-scripts/make-sourceball 2009-03-12 17:54:35.000000000 -0400 +++ make-sourceball 2009-08-14 22:54:59.000000000 -0400 @@ -63,7 +63,7 @@ done done - return 1 + return 0 } create_srcpackage() {
On Fri, Aug 14, 2009 at 10:09 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
On Fri, Aug 14, 2009 at 10:26 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
On Fri, Aug 14, 2009 at 9:44 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
Another change, will fix commit: make-sourceball: license skip is NOT a failure http://projects.archlinux.org/?p=dbscripts.git;a=commit;h=891e4f353d098a639c...
This patch will remove skip licences from failed.txt
--- /arch/misc-scripts/make-sourceball 2009-03-12 17:54:35.000000000 -0400 +++ make-sourceball 2009-08-14 22:54:59.000000000 -0400 @@ -63,7 +63,7 @@ done done
- return 1 + return 0 }
create_srcpackage() {
Hmm this change makes it so the chk_license function never fails, and it needs to. Otherwise we get tarballs for every package regardless of license
On Tue, Aug 18, 2009 at 6:57 PM, Aaron Griffin<aaronmgriffin@gmail.com> wrote:
On Fri, Aug 14, 2009 at 10:09 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
On Fri, Aug 14, 2009 at 10:26 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
On Fri, Aug 14, 2009 at 9:44 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
Another change, will fix commit: make-sourceball: license skip is NOT a failure http://projects.archlinux.org/?p=dbscripts.git;a=commit;h=891e4f353d098a639c...
This patch will remove skip licences from failed.txt
--- /arch/misc-scripts/make-sourceball 2009-03-12 17:54:35.000000000 -0400 +++ make-sourceball 2009-08-14 22:54:59.000000000 -0400 @@ -63,7 +63,7 @@ done done
- return 1 + return 0 }
create_srcpackage() {
Hmm this change makes it so the chk_license function never fails, and it needs to. Otherwise we get tarballs for every package regardless of license
The problem that I wanted to fix was that failed.txt is cluttered by packages that have licenses for which we don't want to make tarballs for (GPL3, BSD MIT, etc). I don't think these should be treated as failures as we expect the script to not make tarballs for them. I guess I'll need to come up with a better implementation unless you want these packages in failed.txt
On Fri, Aug 14, 2009 at 8:44 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
Hi,
Here is a patch (only provided inline) to fix any arch support to the dbscripts. It will add any support for the sourceball script and probably fix the clean up scripts. It also set the BUILDSCRIPT as it will be no longer defined in makepkg.conf as of pacman 3.3. The patch for db-update is required as the any arch is treated as a special case (and db-update works with current /arch/config ).
Eric
--- /arch/config 2009-08-03 12:31:28.000000000 -0400 +++ config 2009-08-14 18:32:02.000000000 -0400 @@ -8,6 +8,5 @@
STAGING="$HOME/staging" TMPDIR="/srv/tmp" -ARCHES=(i686 x86_64) -# To be used if $BUILDSCRIPT is removed from makepkg.conf -#BUILDSCRIPT="PKGBUILD" +ARCHES=(i686 x86_64 any) +BUILDSCRIPT="PKGBUILD"
--- /arch/db-update 2009-08-03 12:31:28.000000000 -0400 +++ db-update 2009-08-14 19:59:27.000000000 -0400 @@ -106,6 +106,7 @@
for current_arch in ${ARCHES[@]}; do
+ if [ "$current_arch" != "any" ]; then ftppath="$FTP_BASE/$reponame/os/$current_arch" ftppath_any="$FTP_BASE/$reponame/os/any"
@@ -226,6 +227,7 @@ fi
repo_unlock $reponame $current_arch + fi done
if [ -n "$to_add_any" ]; then
Hmmm, In theory this shouldn't be needed, right? The db-update script processes the any arch separately, and the sourceballs script runs through actual package files There may be issues with -f checks and the like and symlinks, but I'd rather solve it that way. Considering it just loops through every package file, it should pick up the symlinks too. If it's not picking those up, let's try to fix it there I just pushed a handful of changes to git, including a change to support split packages here... using pkgbase instead of pkgname. Untested, of course
On Tue, Aug 18, 2009 at 7:22 PM, Aaron Griffin<aaronmgriffin@gmail.com> wrote:
On Fri, Aug 14, 2009 at 8:44 PM, Eric Bélanger<snowmaniscool@gmail.com> wrote:
Hi,
Here is a patch (only provided inline) to fix any arch support to the dbscripts. It will add any support for the sourceball script and probably fix the clean up scripts. It also set the BUILDSCRIPT as it will be no longer defined in makepkg.conf as of pacman 3.3. The patch for db-update is required as the any arch is treated as a special case (and db-update works with current /arch/config ).
Eric
--- /arch/config 2009-08-03 12:31:28.000000000 -0400 +++ config 2009-08-14 18:32:02.000000000 -0400 @@ -8,6 +8,5 @@
STAGING="$HOME/staging" TMPDIR="/srv/tmp" -ARCHES=(i686 x86_64) -# To be used if $BUILDSCRIPT is removed from makepkg.conf -#BUILDSCRIPT="PKGBUILD" +ARCHES=(i686 x86_64 any) +BUILDSCRIPT="PKGBUILD"
--- /arch/db-update 2009-08-03 12:31:28.000000000 -0400 +++ db-update 2009-08-14 19:59:27.000000000 -0400 @@ -106,6 +106,7 @@
for current_arch in ${ARCHES[@]}; do
+ if [ "$current_arch" != "any" ]; then ftppath="$FTP_BASE/$reponame/os/$current_arch" ftppath_any="$FTP_BASE/$reponame/os/any"
@@ -226,6 +227,7 @@ fi
repo_unlock $reponame $current_arch + fi done
if [ -n "$to_add_any" ]; then
Hmmm, In theory this shouldn't be needed, right? The db-update script processes the any arch separately, and the sourceballs script runs through actual package files
There may be issues with -f checks and the like and symlinks, but I'd rather solve it that way. Considering it just loops through every package file, it should pick up the symlinks too. If it's not picking those up, let's try to fix it there
I just pushed a handful of changes to git, including a change to support split packages here... using pkgbase instead of pkgname. Untested, of course
I think it's needed. The sourceball script use the arch to get the sourceball name: for arch in ${ARCHES[@]}; do ... srcpkg="${srcpkg//-$arch/}" About db-update, you could go throught the loops and add tests so that adding any to ARCHES doen't break anything. AFAICS, SRCEXT is no longer defined. You should add it to /arch/config
participants (2)
-
Aaron Griffin
-
Eric Bélanger