Bug#906643: Bug#911832: Bug#906643: transition: php7.3
Paul, thank you for all the effort you have put into this. I understand and appreciate the effort. I am basically not objecting to whatever solution there will be, it is ultimately that package maintainer’s decision, and I am sorry I can’t be more helpful, but my free time is limited and all I wanted was to clear the packages preventing faster migration of php-defaults to testing (which I failed anyway because of that last package...) Thanks, Ondřej -- Ondřej Surý > On 9 Nov 2018, at 21:12, Paul Gevers wrote: > > Hi, > >> On 09-11-18 20:09, Ondřej Surý wrote: >> But somehow for this simple package I would just prefer to just bite >> the bullet once in a while, do binNMU and then suffer it through… My >> experience tells me that “the simpler the better” even if it’s not >> perfect. Perfect but complex tend to break in more mysterious way... > > Yes, except that for regressions there may be more people watch (like > me) that will not know these details and waste their time understanding. > > I've just triggered the combination now, but I am not happy yet with the > final outcome. > > Paul >
Bug#906643: Bug#911832: Bug#906643: transition: php7.3
Hi, On 09-11-18 20:09, Ondřej Surý wrote: > But somehow for this simple package I would just prefer to just bite > the bullet once in a while, do binNMU and then suffer it through… My > experience tells me that “the simpler the better” even if it’s not > perfect. Perfect but complex tend to break in more mysterious way... Yes, except that for regressions there may be more people watch (like me) that will not know these details and waste their time understanding. I've just triggered the combination now, but I am not happy yet with the final outcome. Paul signature.asc Description: OpenPGP digital signature
Bug#906643: Bug#911832: Bug#906643: transition: php7.3
Well, it’s either making the package binNMUable (using generic php-fpm test dependency) or recording the exact dependencies (using php7.3-fpm) or dynamically generating debian/tests/control like I do for php-defaults (which requires much more complex system). But somehow for this simple package I would just prefer to just bite the bullet once in a while, do binNMU and then suffer it through… My experience tells me that “the simpler the better” even if it’s not perfect. Perfect but complex tend to break in more mysterious way... If you can come up with something smarter, then I am sure the maintainer of the package would be all ears. Ondrej -- Ondřej Surý ond...@sury.org > On 9 Nov 2018, at 19:33, Paul Gevers wrote: > > Hi Ondřej, > > On 09-11-18 18:39, OndÅej Surý wrote: >> The versioned depends is needed only for autopkgtests and not for the >> package itself. So, I think the dependency is useless there. > > I misunderstood you then. Still, if a test case of a package requires a > different specific (minimum) version of some other package than the > package itself, the debian/tests/control file could (and in my opinion > should) document that. How could our migration software add the right > triggers otherwise? > > Paul > >> >> Ondrej >> -- >> Ondřej Surý >> >>> On 9 Nov 2018, at 10:37, Paul Gevers wrote: >>> >>> Hi, >>> >>> Hmm, I should read my backlog before replying. >>> On 08-11-18 22:53, Ondřej Surý wrote: But php-defaults and rss-bridge needs to go together. >>> >>> That is ok, but where is this coded in the dependencies? >>> I thought that runtime detection of default PHP version in autopkgtest would be overkill, so the socket path is hardcoded at the build-time. >>> >>> I don't consider runtime detection an issue here. The issue is that the >>> dependency system isn't notified of the relation AFAICT. Options I see >>> are that either rss-bridge needs to tell which version of php it needs, >>> or php-defaults needs to add a versioned breaks on rss-bridge. >>> >>> Paul >>> >> >
Bug#906643: Bug#911832: Bug#906643: transition: php7.3
Hi Ondřej, On 09-11-18 18:39, OndÅej Surý wrote: > The versioned depends is needed only for autopkgtests and not for the package > itself. So, I think the dependency is useless there. I misunderstood you then. Still, if a test case of a package requires a different specific (minimum) version of some other package than the package itself, the debian/tests/control file could (and in my opinion should) document that. How could our migration software add the right triggers otherwise? Paul > > Ondrej > -- > Ondřej Surý > >> On 9 Nov 2018, at 10:37, Paul Gevers wrote: >> >> Hi, >> >> Hmm, I should read my backlog before replying. >> >>> On 08-11-18 22:53, Ondřej Surý wrote: >>> But php-defaults and rss-bridge needs to go together. >> >> That is ok, but where is this coded in the dependencies? >> >>> I thought that runtime detection of default PHP version in autopkgtest >>> would be overkill, so the socket path is hardcoded at the build-time. >> >> I don't consider runtime detection an issue here. The issue is that the >> dependency system isn't notified of the relation AFAICT. Options I see >> are that either rss-bridge needs to tell which version of php it needs, >> or php-defaults needs to add a versioned breaks on rss-bridge. >> >> Paul >> > signature.asc Description: OpenPGP digital signature
Bug#906643: Bug#911832: Bug#906643: transition: php7.3
The versioned depends is needed only for autopkgtests and not for the package itself. So, I think the dependency is useless there. Ondrej -- Ondřej Surý > On 9 Nov 2018, at 10:37, Paul Gevers wrote: > > Hi, > > Hmm, I should read my backlog before replying. > >> On 08-11-18 22:53, Ondřej Surý wrote: >> But php-defaults and rss-bridge needs to go together. > > That is ok, but where is this coded in the dependencies? > >> I thought that runtime detection of default PHP version in autopkgtest would >> be overkill, so the socket path is hardcoded at the build-time. > > I don't consider runtime detection an issue here. The issue is that the > dependency system isn't notified of the relation AFAICT. Options I see > are that either rss-bridge needs to tell which version of php it needs, > or php-defaults needs to add a versioned breaks on rss-bridge. > > Paul >
Bug#906643: Bug#911832: Bug#906643: transition: php7.3
Hi, Hmm, I should read my backlog before replying. On 08-11-18 22:53, Ondřej Surý wrote: > But php-defaults and rss-bridge needs to go together. That is ok, but where is this coded in the dependencies? > I thought that runtime detection of default PHP version in autopkgtest would > be overkill, so the socket path is hardcoded at the build-time. I don't consider runtime detection an issue here. The issue is that the dependency system isn't notified of the relation AFAICT. Options I see are that either rss-bridge needs to tell which version of php it needs, or php-defaults needs to add a versioned breaks on rss-bridge. Paul signature.asc Description: OpenPGP digital signature
Bug#906643: Bug#911832: Bug#906643: transition: php7.3
Hi Ondřej, On 08-11-18 22:47, Ondřej Surý wrote: >> On 9 Nov 2018, at 03:35, Paul Gevers wrote: >> >> You also uploaded (NMU) two revisions of rss-bridge. The last one is >> stuck in unstable because you broke the autopkgtest. > > Umm, no? > > https://ci.debian.net/data/autopkgtest/unstable/amd64/r/rss-bridge/1281281/log.gz > https://ci.debian.net/packages/r/rss-bridge/unstable/amd64/ I should have phrased that differently. Your latest version fails it's autopkgtest when that version is installed in testing. https://ci.debian.net/packages/r/rss-bridge/testing/amd64/ https://qa.debian.org/excuses.php?package=rss-bridge As it passes in unstable, it is probably missing a versioned (test) dependency somewhere. > Also php-doctrine-data-fixture haven’t been run with php-doctrine-orm >= > 2.6.2+dfsg-2 yet in unstable. I just triggered that. > In testing ci, it passes now: > https://ci.debian.net/data/autopkgtest/testing/amd64/p/php-doctrine-data-fixtures/1284518/log.gz But in testing php-doctrine-data-fixtures was already passing. So it is good that updating doctrine still passed, but what we want to see if it passes with php-defaults/68. I'll trigger that combination shortly. Paul signature.asc Description: OpenPGP digital signature
Bug#906643: Bug#911832: Bug#906643: transition: php7.3
But php-defaults and rss-bridge needs to go together. I thought that runtime detection of default PHP version in autopkgtest would be overkill, so the socket path is hardcoded at the build-time. Ondrej -- Ondřej Surý ond...@sury.org > On 9 Nov 2018, at 04:47, Ondřej Surý wrote: > > Hi Paul, > > >> On 9 Nov 2018, at 03:35, Paul Gevers wrote: >> >> You also uploaded (NMU) two revisions of rss-bridge. The last one is >> stuck in unstable because you broke the autopkgtest. > > Umm, no? > > https://ci.debian.net/data/autopkgtest/unstable/amd64/r/rss-bridge/1281281/log.gz > https://ci.debian.net/packages/r/rss-bridge/unstable/amd64/ > > I made the package binNMUable… > > Also php-doctrine-data-fixture haven’t been run with php-doctrine-orm >= > 2.6.2+dfsg-2 yet in unstable. > > In testing ci, it passes now: > https://ci.debian.net/data/autopkgtest/testing/amd64/p/php-doctrine-data-fixtures/1284518/log.gz > > O. > -- > Ondřej Surý > ond...@sury.org
Bug#906643: Bug#911832: Bug#906643: transition: php7.3
Hi Paul, > On 9 Nov 2018, at 03:35, Paul Gevers wrote: > > You also uploaded (NMU) two revisions of rss-bridge. The last one is > stuck in unstable because you broke the autopkgtest. Umm, no? https://ci.debian.net/data/autopkgtest/unstable/amd64/r/rss-bridge/1281281/log.gz https://ci.debian.net/packages/r/rss-bridge/unstable/amd64/ I made the package binNMUable… Also php-doctrine-data-fixture haven’t been run with php-doctrine-orm >= 2.6.2+dfsg-2 yet in unstable. In testing ci, it passes now: https://ci.debian.net/data/autopkgtest/testing/amd64/p/php-doctrine-data-fixtures/1284518/log.gz O. -- Ondřej Surý ond...@sury.org
Bug#906643: Bug#911832: Bug#906643: transition: php7.3
Hi Ondřej, On 07-11-18 20:48, Ondřej Surý wrote: > I solved the doctrine bug, but php-symfony-polyfill 1.10.0 turned out to be > harder nut to crack: > > For reference: > > 1. I removed references for Normalizer::NONE as they were testing if the code > would "assert" (whatever that means in this context) > 2. There was mismatch between declaration of warning() function in > TestListener > 3. I commented out all IDNA2003 tests as I suspect that something has moved > to IDNA2008 and that's causing failure > > And then it started to be beyond my current understanding of the code and the > amount of time I would like to spend fixing it. You also uploaded (NMU) two revisions of rss-bridge. The last one is stuck in unstable because you broke the autopkgtest. On top of that, the autopkgtest in testing (your first version) still fails with php-defaults, but passes otherwise. Paul signature.asc Description: OpenPGP digital signature