0001111: Provide source tarballs for stable releases

(0003086)
jeanphi (administrator)
2011-07-01 08:31

The MD5 sum of the generated tarball shouldn't change at all, the content being the same. Would you have the ability to make a diff of the download in case the MD5 sum changes? I can only see it changing because of serveur option modifications regarding compression, but that is pretty unusual.

(0003088)
stativ (reporter)
2011-07-01 11:00

jeanphi: I'll do that. It will probably take some time, because it usually took a few days for the check sum to change.

(0003090)
jeanphi (administrator)
2011-07-04 05:32

Thanks a lot, I'm waiting for some more news from you.

(0003094)
stativ (reporter)
2011-07-09 13:09

I just don't get it. According to the logs we had to rebuild luxrender package for archlinux three times in about two weeks because md5sum of the source tarball changed. But now there is no change for nine days.

First we downloaded sources from bitbucket, but the md5sum changed, so we started to use luxrender.net mirror with hope that the md5sum won't change. But it changed twice after that. From bugzilla and IRC log the md5sum changed approximately 25th June and 30th June.

(0003095)
J the Ninja (developer)
2011-07-09 14:13

While the v08 tag didn't change in that time, the repo as a whole did (new commits to the trunk). Could Mercurial be changing something in the contents of the tagged tarball because of a new commit? Thus making the md5sum change?

(0003098)
jeanphi (administrator)
2011-07-15 02:32

I don't think so, and there have been lots of changes since without the MD5 changing.

(0003100)
stativ (reporter)
2011-07-15 10:04

Is it possible that the server was updated at that time?

But then it would be strange coincidence that md5sums of tarball from bitbucket changed at least once too at that time…

Maybe this issue should be closed so it doesn't haunt the bugzilla. I can reopen it (or open another one if reopening is not possible in mantis) if I find more information about this issue if it ever happens again.

Issue History
Date Modified	Username	Field	Change
2011-06-30 02:42	stativ	New Issue
2011-07-01 08:31	jeanphi	Note Added: 0003086
2011-07-01 08:31	jeanphi	Assigned To	=> jeanphi
2011-07-01 08:31	jeanphi	Status	new => feedback
2011-07-01 11:00	stativ	Note Added: 0003088
2011-07-01 11:00	stativ	Status	feedback => assigned
2011-07-04 05:32	jeanphi	Note Added: 0003090
2011-07-09 13:09	stativ	Note Added: 0003094
2011-07-09 14:13	J the Ninja	Note Added: 0003095
2011-07-15 02:32	jeanphi	Note Added: 0003098
2011-07-15 10:04	stativ	Note Added: 0003100
2011-07-19 09:37	jeanphi	Status	assigned => resolved
2011-07-19 09:37	jeanphi	Fixed in Version	=> 1.0
2011-07-19 09:37	jeanphi	Resolution	open => unable to reproduce
2011-08-31 05:17	jeanphi	Target Version	=> 1.0
2012-08-21 06:54	jeanphi	Fixed in Version	1.0 => 1.0RC4
2012-09-11 06:10	jeanphi	Status	resolved => closed

Notes
(0003086) jeanphi (administrator) 2011-07-01 08:31	The MD5 sum of the generated tarball shouldn't change at all, the content being the same. Would you have the ability to make a diff of the download in case the MD5 sum changes? I can only see it changing because of serveur option modifications regarding compression, but that is pretty unusual.

(0003088) stativ (reporter) 2011-07-01 11:00	jeanphi: I'll do that. It will probably take some time, because it usually took a few days for the check sum to change.

(0003090) jeanphi (administrator) 2011-07-04 05:32	Thanks a lot, I'm waiting for some more news from you.

(0003094) stativ (reporter) 2011-07-09 13:09	I just don't get it. According to the logs we had to rebuild luxrender package for archlinux three times in about two weeks because md5sum of the source tarball changed. But now there is no change for nine days. First we downloaded sources from bitbucket, but the md5sum changed, so we started to use luxrender.net mirror with hope that the md5sum won't change. But it changed twice after that. From bugzilla and IRC log the md5sum changed approximately 25th June and 30th June.

(0003095) J the Ninja (developer) 2011-07-09 14:13	While the v08 tag didn't change in that time, the repo as a whole did (new commits to the trunk). Could Mercurial be changing something in the contents of the tagged tarball because of a new commit? Thus making the md5sum change?

(0003098) jeanphi (administrator) 2011-07-15 02:32	I don't think so, and there have been lots of changes since without the MD5 changing.

(0003100) stativ (reporter) 2011-07-15 10:04	Is it possible that the server was updated at that time? But then it would be strange coincidence that md5sums of tarball from bitbucket changed at least once too at that time… Maybe this issue should be closed so it doesn't haunt the bugzilla. I can reopen it (or open another one if reopening is not possible in mantis) if I find more information about this issue if it ever happens again.

Relationships