[pacman-dev] 3.2.1 release
While there is nothing critical, all bugfixes already fixed in git are nice to have, but there are especially two that keep getting reported : makepkg: fix download functions with weird urls http://projects.archlinux.org/?p=pacman.git;a=commit;h=9bc799ec7b1718e6d90ce... Fix creation of source package with local source files http://projects.archlinux.org/?p=pacman.git;a=commit;h=57bd8974c73571d827617... So the sooner, the better :) The only pending patches are the following ones : pacman/callback.c : fix detection of totaldownload http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=commit;h=4f563... Fix error when sourcing profile script http://dev.archlinux.org/~allan/gitweb/gitweb.cgi?p=pacman.git;a=commitdiff;... And eventually this one : Allow sync200 pactest to pass without libdownload. http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=commitdiff;h=6... This last one also shows how to use curl as XferCommand in pacman.conf, which is nice because it handles file:// protocol.
On 8/13/08, Xavier <shiningxc@gmail.com> wrote:
While there is nothing critical, all bugfixes already fixed in git are nice to have, but there are especially two that keep getting reported : So the sooner, the better :)
I'm on a business trip this week so its a bit harder for me to keep on top of this stuff at night. I think it would be wise to release 3.2.1 next week if nothing else critical comes up- this release has been awesome. Looking through what you mentioned: totaldownload patch: definitely include source profile error patch: definitely include libdownload/curl patch: I'd actually like to see this one split up- one patch to change the "default" external downloader to curl which I will definitely include. However, I am hesitant to leave the internal download code completely untested which is what the second half of that patch does. Can anyone think of a way we can do this? And Xavier- I know there are still some patches you have on your working branch I haven't pulled or commented on yet, leaving you a bit out to dry. I will try to do that ASAP on the list here. -Dan
On Wed, Aug 13, 2008 at 9:34 PM, Dan McGee <dpmcgee@gmail.com> wrote:
On 8/13/08, Xavier <shiningxc@gmail.com> wrote:
While there is nothing critical, all bugfixes already fixed in git are nice to have, but there are especially two that keep getting reported : So the sooner, the better :)
I'm on a business trip this week so its a bit harder for me to keep on top of this stuff at night. I think it would be wise to release 3.2.1 next week if nothing else critical comes up- this release has been awesome.
Ok, good.
Looking through what you mentioned: totaldownload patch: definitely include source profile error patch: definitely include libdownload/curl patch: I'd actually like to see this one split up- one patch to change the "default" external downloader to curl which I will definitely include. However, I am hesitant to leave the internal download code completely untested which is what the second half of that patch does. Can anyone think of a way we can do this?
Hmm, by the way, I did not change the "default" XferCommand, I added one. I thought it was alright to give both the wget and curl examples. wget is probably the most widely used one. But curl is likely to be available on the majority of systems, and supports the file:// protocol : Required By : bzflag git gnupg google-earth moc octave3 openoffice-base rtorrent xine-ui About the other half of the patch, I agree it is not so nice. And the only way I can think of is too complicated for my taste. At least, I guess it is possible to use configure and friends to only define that XferCommand when libdownload support is disabled. I guess I will just drop that patch, it does not matter. Besides there are already the 2 scriptlets pactest which fail on cygwin, because fakechroot is missing. So really no big deal.
And Xavier- I know there are still some patches you have on your working branch I haven't pulled or commented on yet, leaving you a bit out to dry. I will try to do that ASAP on the list here.
That would be nice too, but still lower priority so no problem :)
Am Wed, 13 Aug 2008 23:43:02 +0200 schrieb Xavier <shiningxc@gmail.com>:
About the other half of the patch, I agree it is not so nice. And the only way I can think of is too complicated for my taste. At least, I guess it is possible to use configure and friends to only define that XferCommand when libdownload support is disabled. I guess I will just drop that patch, it does not matter. Besides there are already the 2 scriptlets pactest which fail on cygwin, because fakechroot is missing. So really no big deal.
I wrote a Makefile.am for the tests dir, which works very much like the one for scripts/ . If you change the test to something like: if not "download" in "@LIBS@": self.option['XferCommand'] = ['/usr/bin/curl %u > %o'] rename it to sync200.py.in and run automake and friends, both cases should be covered. I'm not sure if this is really worth the trouble, but this could be useful in other cases.
On Fri, Aug 15, 2008 at 12:42 AM, Henning Garus <henning.garus@googlemail.com> wrote:
I wrote a Makefile.am for the tests dir, which works very much like the one for scripts/ . If you change the test to something like:
if not "download" in "@LIBS@": self.option['XferCommand'] = ['/usr/bin/curl %u > %o']
rename it to sync200.py.in and run automake and friends, both cases should be covered. I'm not sure if this is really worth the trouble, but this could be useful in other cases.
Ah yeah, nice job, it works perfectly and it is less complex than I thought. Thanks! I added your patch and the sync200 fix to my working branch : http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=shortlog;h=ref... Now the ball is to Dan to decide whether he wants these or not :)
Update: FS#11203 - pacman doesn't care for NoPassiveFtp anymore Fixed here : http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=commit;h=664da... FS#11218 - pacman gives "No such file" error during some --force installs No real patches yet, but concrete suggestions in the comments FS#11225 - repo-add creates wrong CSIZE entry This was not really a bug because it was intended that way. But maybe it was not a good idea. I like the idea of having a SIZECMD in /etc/makepkg.conf so that any user can specify a command that works on his system. But do we want to fix this for 3.2.1? If yes, we definitely need to hardcode a default SIZECMD inside makepkg, to keep backward compatibility of makepkg.conf
On Mon, Aug 18, 2008 at 10:47 PM, Xavier <shiningxc@gmail.com> wrote:
Update:
FS#11203 - pacman doesn't care for NoPassiveFtp anymore Fixed here : http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=commit;h=664da...
FS#11218 - pacman gives "No such file" error during some --force installs No real patches yet, but concrete suggestions in the comments
A fix on my maint-working branch http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=commit;h=962a9...
FS#11225 - repo-add creates wrong CSIZE entry This was not really a bug because it was intended that way. But maybe it was not a good idea. I like the idea of having a SIZECMD in /etc/makepkg.conf so that any user can specify a command that works on his system. But do we want to fix this for 3.2.1? If yes, we definitely need to hardcode a default SIZECMD inside makepkg, to keep backward compatibility of makepkg.conf
A fix on my working branch http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=commit;h=0edb8...
A fix on my maint-working branch http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=commit;h=962a9...
A fix on my working branch http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=commit;h=0edb8...
Both patches look fine to me. Bye
On Mon, Aug 18, 2008 at 6:16 PM, Nagy Gabor <ngaba@bibl.u-szeged.hu> wrote:
A fix on my maint-working branch http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=commit;h=962a9...
A fix on my working branch http://shining.toofishes.net/gitweb/gitweb.cgi?p=pacman.git;a=commit;h=0edb8...
Both patches look fine to me.
OK, the current 3.2.1 shortlog is below. I think I took care of everything mentioned above with the push tonight except for the csize/stat stuff. I commented on that bug report so you can see my thoughts there. FS#11203 and FS#11218 should be fixed. Anything else that needs to sneak into 3.2.1? Allan McRae (2): Fix creation of source package with local source files Fix error when sourcing profile script Dan McGee (4): makepkg: Clarify usage instructions for --allsource/--source doc: allow asciidoc to format the note Fix variable naming issues in _alpm_db_cmp Fix segfault on x86_64 when using UseSyslog Nagy Gabor (2): Remove an usused variable from alpm/util.c/_alpm_lckmk() repo-add: Fix whitespaces handling in variables. Xavier Chantry (8): makepkg: fix download functions with weird urls. dload.c : drop the specific handling of file: url. Avoid double slashes in URLs given to libdownload. pacman/callback.c : fix detection of totaldownload pacman.conf : add curl example for XferCommand. pacman.c: fix typo s/NoPassiveFTP/NoPassiveFtp libalpm/add.c : ensure the old pkg was fully loaded. repo-add: add optdepends to the sync database.
participants (4)
-
Dan McGee
-
Henning Garus
-
Nagy Gabor
-
Xavier