[aur-general] AUR repos and git subtrees
I have just attempted to update fanficfare{,-git} and I cannot push the changes because they are a non-fast-forward. I cannot be 100% sure because I don't have local clones, I have been splitting off subtrees as needed (see my setup at https://github.com/eli-schwartz/pkgbuilds) but this is supposed to be 100% guaranteed to create the same sha1 hash. However, it appears that the AUR repos are now out of sync. Same content and everything. Am I supposed to believe that those packages -- and a bunch more (but not all...) of mine -- had their repos randomly and sneakily rewritten??? Or that git subtrees are simply unreliable??? Neither seems very likely. This was working perfectly at least until July 20 (the last time I updated anything). -- Eli Schwartz
On 9 August 2015 at 19:08, Eli Schwartz <eschwartz93@gmail.com> wrote:
I have just attempted to update fanficfare{,-git} and I cannot push the changes because they are a non-fast-forward.
I cannot be 100% sure because I don't have local clones, I have been splitting off subtrees as needed (see my setup at https://github.com/eli-schwartz/pkgbuilds) but this is supposed to be 100% guaranteed to create the same sha1 hash. However, it appears that the AUR repos are now out of sync. Same content and everything.
Am I supposed to believe that those packages -- and a bunch more (but not all...) of mine -- had their repos randomly and sneakily rewritten??? Or that git subtrees are simply unreliable??? Neither seems very likely.
This was working perfectly at least until July 20 (the last time I updated anything).
-- Eli Schwartz
I know it makes no sense why it should, but maybe your commit 84953afaf0255b746b3f591cf869d530f31b3a09 made it confused? Whenever something goes wrong for me with git I always try out a fresh clone to mess around with so you might want to see if that works as well. I unfortunately did notice that clones don't retain their subtree infrastructure though, so that's kind of disappointing (I had to merge your commit manually). Thanks for fanficfare though, didn't know that existed. Sincerely, Joost Bremmer "We apologize for the inconvenience"
On Sun, Aug 9, 2015 at 8:00 PM, Joost Bremmer <toost.b@gmail.com> wrote:
I know it makes no sense why it should, but maybe your commit 84953afaf0255b746b3f591cf869d530f31b3a09 made it confused? Whenever something goes wrong for me with git I always try out a fresh clone to mess around with so you might want to see if that works as well. I unfortunately did notice that clones don't retain their subtree infrastructure though, so that's kind of disappointing (I had to merge your commit manually).
Thanks for fanficfare though, didn't know that existed.
Sincerely, Joost Bremmer
"We apologize for the inconvenience"
Nope, a clean clone from an earlier history does exactly the same thing. As for my subtrees -- I don't rejoin them. Maybe if I had, I would either know something weird was going on, or I wouldn't end up with new sha1 histories... Glad you enjoy fanficfare though. :) I pathologically haunt the Mobileread.com forums, so I know a little *too* much about these things. :rofl: (Do note there is a calibre plugin as well, which offers enhanced functionality). -- Eli Schwartz
On Sun, Aug 9, 2015 at 1:08 PM, Eli Schwartz <eschwartz93@gmail.com> wrote:
I have just attempted to update fanficfare{,-git} and I cannot push the changes because they are a non-fast-forward.
I cannot be 100% sure because I don't have local clones, I have been splitting off subtrees as needed (see my setup at https://github.com/eli-schwartz/pkgbuilds) but this is supposed to be 100% guaranteed to create the same sha1 hash. However, it appears that the AUR repos are now out of sync. Same content and everything.
Am I supposed to believe that those packages -- and a bunch more (but not all...) of mine -- had their repos randomly and sneakily rewritten??? Or that git subtrees are simply unreliable??? Neither seems very likely.
This was working perfectly at least until July 20 (the last time I updated anything).
-- Eli Schwartz
So, comparing my subtree split of fanficfare I get a root commit at: 69c3f0eec7fe3ce2983dec09b5dcba5a02d9169c In the AUR, the root commit is: 3c5d51989c6d14feea568ba6627dd9e5e7492f67 git diff them, they're identical trees. The commit messages are the same, the author and commit info are identical. How is that even possible??? -- Eli Schwartz
On Mon, 17 Aug 2015 at 21:09:28, Eli Schwartz wrote:
[...] So, comparing my subtree split of fanficfare I get a root commit at: 69c3f0eec7fe3ce2983dec09b5dcba5a02d9169c In the AUR, the root commit is: 3c5d51989c6d14feea568ba6627dd9e5e7492f67
git diff them, they're identical trees. The commit messages are the same, the author and commit info are identical. [...]
And both commits do not have any ancestors? If you execute `git cat-file commit <id>` (using 69c3f0e, respectively 3c5d519 as commit hashes), are the outputs exactly the same (diff(1) them)?
On Mon, Aug 17, 2015 at 3:25 PM, Lukas Fleischer <lfleischer@archlinux.org> wrote:
And both commits do not have any ancestors?
If you execute `git cat-file commit <id>` (using 69c3f0e, respectively 3c5d519 as commit hashes), are the outputs exactly the same (diff(1) them)?
Thanks for the suggestion! No, there aren't any ancestors, but there is this... ``` tree 0a0dd290957d796d5aa3d83c830ed8094b48e7a3 -author Eli Schwartz <eschwartz93@gmail.com> 1429295767 -0400 -committer Eli Schwartz <eschwartz93@gmail.com> 1433830927 -0400 +author Eli Schwartz <eschwartz93@gmail.com> 1429295767 -0500 +committer Eli Schwartz <eschwartz93@gmail.com> 1433830927 -0500 Initial upload: fanficfare 2.2.2-1 ``` And the problem goes away when I stop using git config log.date local Problem solved, seems to be a bug in git. -- Eli Schwartz
participants (3)
-
Eli Schwartz
-
Joost Bremmer
-
Lukas Fleischer