[aur-general] perl prereleases request to make official decision to deviate from general guidleines
I'm working on a PKGBUILD for perl 5.12 and I notice that it's providing 20 some odd prerelease modules. I'm annoyed but I can't do anything about it. The general PKGBUILD guidelines say use the same as upstream unless there is a hyphen. so for a perl prerelease of 0.90_1 that guideline follows to release it as exactly the same. However 0.90_1 is == 0.90.1 and greater than 0.90 which it is a prerelease for on cpan. I'd like to suggest that for perl the official guideline be to make the version 0.90_pre1 Dist::Zilla a release automation tool also allows for using a format _TRIAL which is not yet popular on cpan but does happen should pacman add a _trial ? to the list of options? I think we could just use _pre here too. -- Caleb Cushing http://xenoterracide.blogspot.com
I'm working on a PKGBUILD for perl 5.12 and I notice that it's providing 20 some odd prerelease modules. I'm annoyed but I can't do anything about it.
The general PKGBUILD guidelines say use the same as upstream unless there is a hyphen. so for a perl prerelease of 0.90_1 that guideline follows to release it as exactly the same. However 0.90_1 is == 0.90.1 and greater than 0.90 which it is a prerelease for on cpan.
I'd like to suggest that for perl the official guideline be to make the version 0.90_pre1
Dist::Zilla a release automation tool also allows for using a format _TRIAL which is not yet popular on cpan but does happen should pacman add a _trial ? to the list of options? I think we could just use _pre here too. -- Caleb Cushing
See my previous reply to the thread PKGVER (prerelease hell). I agree that Arch's CPAN version scheme is flawed (mixing of old and new CPAN schemes, failure to understand CPAN schemes (e.g. 0.30 is actually 0.300), naïve handling of release versions, etc). I believe the formatting function that I've written to convert CPAN versions to Pacman versions addresses these issues. If you agree and decide to use it too, I think we could push harder for CPAN package uniformity in Arch to try to clean up the current mess, which is an unnecessary source of frustration. Xyne
On 01/05/10 13:24, Xyne wrote:
I'm working on a PKGBUILD for perl 5.12 and I notice that it's providing 20 some odd prerelease modules. I'm annoyed but I can't do anything about it.
The general PKGBUILD guidelines say use the same as upstream unless there is a hyphen. so for a perl prerelease of 0.90_1 that guideline follows to release it as exactly the same. However 0.90_1 is == 0.90.1 and greater than 0.90 which it is a prerelease for on cpan.
I'd like to suggest that for perl the official guideline be to make the version 0.90_pre1
Dist::Zilla a release automation tool also allows for using a format _TRIAL which is not yet popular on cpan but does happen should pacman add a _trial ? to the list of options? I think we could just use _pre here too. -- Caleb Cushing
See my previous reply to the thread PKGVER (prerelease hell). I agree that Arch's CPAN version scheme is flawed (mixing of old and new CPAN schemes, failure to understand CPAN schemes (e.g. 0.30 is actually 0.300), naïve handling of release versions, etc).
I believe the formatting function that I've written to convert CPAN versions to Pacman versions addresses these issues. If you agree and decide to use it too, I think we could push harder for CPAN package uniformity in Arch to try to clean up the current mess, which is an unnecessary source of frustration.
The easiest way to clean things up would be to package perl yourselves. There is an advertisement about junior developers on the main page so... Obviously given nothing official has been done towards perl 5.12, that might be an area we would consider.
On Fri, Apr 30, 2010 at 11:31 PM, Allan McRae <allan@archlinux.org> wrote:
Obviously given nothing official has been done towards perl 5.12, that might be an area we would consider.
kinda sad that you were able to roll out a new gcc faster than a new perl... anyways... I just finished rolling a PKGBUILD based on the 5.10.1 one with a full new provides list. I got this whole pre thing screwed up ... MY BAD. obviously a new perl means building other things... but let me know what you think. -- Caleb Cushing http://xenoterracide.blogspot.com
On Sat, May 1, 2010 at 4:54 AM, Caleb Cushing <xenoterracide@gmail.com> wrote:
I just finished rolling a PKGBUILD based on the 5.10.1 one with a full new provides list.
oh and the script used to generate said provides in case anyone thinks that could help or be improved. -- Caleb Cushing http://xenoterracide.blogspot.com
On Sat, May 1, 2010 at 4:54 AM, Caleb Cushing <xenoterracide@gmail.com> wrote:
I just finished rolling a PKGBUILD based on the 5.10.1 one with a full new provides list.
I won't actually say that I'm sure this won't eat your computer -- Caleb Cushing http://xenoterracide.blogspot.com
On Sat, May 1, 2010 at 4:54 AM, Caleb Cushing <xenoterracide@gmail.com> wrote:
I just finished rolling a PKGBUILD based on the 5.10.1
So it builds successfully. But doesn't pass it's test suite. Some things are obviously broken in the final build. http://gist.github.com/386418 got the code that I've updated (very little) there. I'm going to stop looking at it for now. 'till I get my new hard drives this next week. If anyone has any idea's on the breakages... -- Caleb Cushing http://xenoterracide.blogspot.com
On 01/05/2010 05:31, Allan McRae wrote:
On 01/05/10 13:24, Xyne wrote:
I'm working on a PKGBUILD for perl 5.12 and I notice that it's providing 20 some odd prerelease modules. I'm annoyed but I can't do anything about it.
The general PKGBUILD guidelines say use the same as upstream unless there is a hyphen. so for a perl prerelease of 0.90_1 that guideline follows to release it as exactly the same. However 0.90_1 is == 0.90.1 and greater than 0.90 which it is a prerelease for on cpan.
I'd like to suggest that for perl the official guideline be to make the version 0.90_pre1
Dist::Zilla a release automation tool also allows for using a format _TRIAL which is not yet popular on cpan but does happen should pacman add a _trial ? to the list of options? I think we could just use _pre here too. -- Caleb Cushing
See my previous reply to the thread PKGVER (prerelease hell). I agree that Arch's CPAN version scheme is flawed (mixing of old and new CPAN schemes, failure to understand CPAN schemes (e.g. 0.30 is actually 0.300), naïve handling of release versions, etc).
I believe the formatting function that I've written to convert CPAN versions to Pacman versions addresses these issues. If you agree and decide to use it too, I think we could push harder for CPAN package uniformity in Arch to try to clean up the current mess, which is an unnecessary source of frustration.
The easiest way to clean things up would be to package perl yourselves. There is an advertisement about junior developers on the main page so... Obviously given nothing official has been done towards perl 5.12, that might be an area we would consider.
I have also written a first draft of a PKGBUILD for perl 5.12.0 already and sent it to Kevin, who is working on improving it. He will release it in testing when it's ready. There are however problems with a bunch of failing test files, which looks like upstream issues to me, so it might take some time before 5.12.0 hits the core repo. That said, I would welcome any applicant interested in packaging Perl and related stuff! @Caleb: I can send you my PKGBUILD if you're interested. Contact kevin@archlinux.org if you have any further suggestions. F
participants (4)
-
Allan McRae
-
Caleb Cushing
-
Firmicus
-
Xyne