[pacman-dev] 3.2 "RC" release
I rebuilt pacman-git tonight with the latest head commit from the GIT repo: http://dev.archlinux.org/~dan/pacman-git/ If anyone notices anything wrong, speak soon. This should be 3.2.0 if nothing comes up, minus the version number bumping. -Dan
On Wed, Jul 30, 2008 at 5:46 AM, Dan McGee <dpmcgee@gmail.com> wrote:
I rebuilt pacman-git tonight with the latest head commit from the GIT repo:
http://dev.archlinux.org/~dan/pacman-git/
If anyone notices anything wrong, speak soon. This should be 3.2.0 if nothing comes up, minus the version number bumping.
You missed one thing about the NEWS file : http://www.archlinux.org/pipermail/pacman-dev/2008-July/012639.html But no problem, you did very well compared to the mass of mails you had to deal with :) I liked all changes you made to all these recent commits (documentation updates for README and submitting-patches, comment for AND-based searching, the STRIP_DIR fix I was supposed to do and didn't do, etc...) So I just pushed the NEWS update to a 32-release branch, as well as all the translation updates I was talking about earlier. One last problem : the pacman.conf needs to be updated. Most important points : 1) remove unstable mirror section 2) add SyncFirst However, I would suggest merging the two versions (package one and git repo one) as much as possible. Attached a proposal. And another trouble I have, but it might be just me, I have no idea, but I no longer get the localized messages. I always get pacman and makepkg in english. I don't remember if it worked with pacman-git before. I tried to disable the debug (DEBUG=0 in PKGBUILD) but it did not help.
One last problem : the pacman.conf needs to be updated. Most important points : 1) remove unstable mirror section 2) add SyncFirst
This may be an overkill, but this can be mentioned in .INSTALL file too, because user may omit pacman.conf.pacnew. Or give a general message: "New options were added to pacman.conf, please check pacman.conf.pacnew (if it exists) and pacman.conf manual for details" Bye
On Wed, Jul 30, 2008 at 12:14 PM, Nagy Gabor <ngaba@bibl.u-szeged.hu> wrote:
One last problem : the pacman.conf needs to be updated. Most important points : 1) remove unstable mirror section 2) add SyncFirst
This may be an overkill, but this can be mentioned in .INSTALL file too, because user may omit pacman.conf.pacnew. Or give a general message: "New options were added to pacman.conf, please check pacman.conf.pacnew (if it exists) and pacman.conf manual for details"
Well I am not sure Dan would like that after what he said here : http://www.archlinux.org/pipermail/arch-dev-public/2008-July/007136.html The impression I got from this is to try to limit the install messages as much as possible, only keeping the critical stuff that would prevent pacman from functioning properly. This is not the case here, pacman 3.2 still works properly with the old config. Then it is the user's duty to make sure all .pacnew files are merged. Ideally, just after an upgrade. But running pacdiff regularly to make sure nothing was missed does not hurt either.
On Wed, Jul 30, 2008 at 12:14 PM, Nagy Gabor <ngaba@bibl.u-szeged.hu> wrote:
One last problem : the pacman.conf needs to be updated. Most important points : 1) remove unstable mirror section 2) add SyncFirst
This may be an overkill, but this can be mentioned in .INSTALL file too, because user may omit pacman.conf.pacnew. Or give a general message: "New options were added to pacman.conf, please check pacman.conf.pacnew (if it exists) and pacman.conf manual for details"
Well I am not sure Dan would like that after what he said here : http://www.archlinux.org/pipermail/arch-dev-public/2008-July/007136.html
I have a completely different opinion. IMHO the current OptDepends is totally _useless_: who checks them with -Qi or -Si?! Personally I have no clue which packages have optional dependencies. I agree with Dan, that optdepends shouldn't be put to install file (they will be "hidden" after package install), but I like that they are listed. I think pacman should automatically print them during package install. (Here we have formatting/localisation difficulties) This behaviour could be disabled (to satisfy Dan ;-). Bye
On Wed, Jul 30, 2008 at 2:12 PM, Nagy Gabor <ngaba@bibl.u-szeged.hu> wrote:
On Wed, Jul 30, 2008 at 12:14 PM, Nagy Gabor <ngaba@bibl.u-szeged.hu> wrote:
One last problem : the pacman.conf needs to be updated. Most important points : 1) remove unstable mirror section 2) add SyncFirst
This may be an overkill, but this can be mentioned in .INSTALL file too, because user may omit pacman.conf.pacnew. Or give a general message: "New options were added to pacman.conf, please check pacman.conf.pacnew (if it exists) and pacman.conf manual for details"
Well I am not sure Dan would like that after what he said here : http://www.archlinux.org/pipermail/arch-dev-public/2008-July/007136.html
I have a completely different opinion. IMHO the current OptDepends is totally _useless_: who checks them with -Qi or -Si?! Personally I have no clue which packages have optional dependencies. I agree with Dan, that optdepends shouldn't be put to install file (they will be "hidden" after package install), but I like that they are listed.
I think pacman should automatically print them during package install. (Here we have formatting/localisation difficulties) This behaviour could be disabled (to satisfy Dan ;-).
I didn't say "don't print anything", I was striking out against the renegade install messages that continued to grow more and more obtrusive. I'm fine with a patch that prints optdepends on package install (and probably also on upgrade if they differ from the previously installed version's list). -Dan
On Wed, Jul 30, 2008 at 9:12 PM, Nagy Gabor <ngaba@bibl.u-szeged.hu> wrote:
Well I am not sure Dan would like that after what he said here : http://www.archlinux.org/pipermail/arch-dev-public/2008-July/007136.html
I have a completely different opinion. IMHO the current OptDepends is totally _useless_: who checks them with -Qi or -Si?! Personally I have no clue which packages have optional dependencies. I agree with Dan, that optdepends shouldn't be put to install file (they will be "hidden" after package install), but I like that they are listed.
I think pacman should automatically print them during package install. (Here we have formatting/localisation difficulties) This behaviour could be disabled (to satisfy Dan ;-).
I think I was not very clear and transformed Dan's words too much. I just talked with him and we agree on these two things : 1) printing optdepends is fine. I actually asked an user to write a feature request : http://bugs.archlinux.org/task/10630 , and I even assigned it to me. But then I ran into annoying problems. Please help, since you care about this issue. 2) messages like "check .pacnew file for new options" are useless. pacman already prints a warning when extracting pacnew files anyway. And you said yourself it might be overkill, so it shouldn't be a problem.
I think I was not very clear and transformed Dan's words too much.
No, I've read that mail, and I reflected to point 2. (which I may misinterpreted: probably Dan don't like them on _upgrade_? Btw, packagers: we have pre/post_install and pre/post_upgrade "scriptlets").
I just talked with him and we agree on these two things : 1) printing optdepends is fine. I actually asked an user to write a feature request : http://bugs.archlinux.org/task/10630 , and I even assigned it to me. But then I ran into annoying problems. Please help, since you care about this issue. Hm. I've missed that FR, thanks.
2) messages like "check .pacnew file for new options" are useless. pacman already prints a warning when extracting pacnew files anyway. And you said yourself it might be overkill, so it shouldn't be a problem.
OK. No problem, I wasn't sure about it neither. I just noted, that some "lazy" users may notice behaviour change (missing SyncFirst option) after upgrade. Bye
On Wed, Jul 30, 2008 at 9:07 AM, Xavier <shiningxc@gmail.com> wrote:
And another trouble I have, but it might be just me, I have no idea, but I no longer get the localized messages. I always get pacman and makepkg in english. I don't remember if it worked with pacman-git before. I tried to disable the debug (DEBUG=0 in PKGBUILD) but it did not help.
This was just some weird kde4 issue : http://bbs.archlinux.org/viewtopic.php?pid=399664#p399664 So you can ignore it.
On Tue, Jul 29, 2008 at 10:46 PM, Dan McGee <dpmcgee@gmail.com> wrote:
I rebuilt pacman-git tonight with the latest head commit from the GIT repo:
http://dev.archlinux.org/~dan/pacman-git/
If anyone notices anything wrong, speak soon. This should be 3.2.0 if nothing comes up, minus the version number bumping.
Dan, you're my hero. I will test this out when I get a chance.
On Wed, Jul 30, 2008 at 5:46 AM, Dan McGee <dpmcgee@gmail.com> wrote:
I rebuilt pacman-git tonight with the latest head commit from the GIT repo:
http://dev.archlinux.org/~dan/pacman-git/
If anyone notices anything wrong, speak soon. This should be 3.2.0 if nothing comes up, minus the version number bumping.
Just noticed one harmless annoyance caused by my SyncFirst patch, but this is a corner case, and really not worth worrying about. I am just mentioning it for the sake of it... $ sudo pacman -Su warning: pacman-git: local (20080730-1) is newer than pacman-git (20080729-1) :: Starting full system upgrade... warning: pacman-git: local (20080730-1) is newer than pacman-git (20080729-1) local database is up to date I installed pacman-git from your repo, then added it to SyncFirst, then upgraded it locally. This message is displayed by _alpm_pkg_compare_versions, so it will be run one first time for the SyncFirst code (used in every -S and -Su operations), then a second time for the -Su code. So this only happens when you have a package in SyncFirst with a local version newer than the repo one. This should be very rare, and then it is just one warning message. So no big deal. And I did not see a very easy quick fix anyway...
participants (4)
-
Aaron Griffin
-
Dan McGee
-
Nagy Gabor
-
Xavier