(per Kyle)
The problem was introduced by this fossil check in (mentioned in the readme.txt):
http://www.fossil-scm.org/index.html/info/bc8d368b66
And as you can see here:
http://www.fossil-scm.org/index.html/timeline?d=bc8d368b&n=126
Version 1.19 is the first version to include it. So, by using the last released version prior to 1.19 (which is 1.18) and patching that the problem goes away.
I don't think I ever investigated the reason that change introduces the problem. I think I most likely ran a bisect and probably used the sqlite_docsrc repository to test with as it has the problem too and I think it's the smallest of the three (fossil, sqlite, sqlite_docsrc).
I see there's a patch to add an incremental export shortly after the problem commit. Makes me wonder if some of those patches could be cherry-picked onto the version that works. :)
Again though, if you use an incremental export and Richard shuns again you can end up with a frankenrepo that no one else can ever reproduce.
(per Kyle)
The problem was introduced by this fossil check in (mentioned in the readme.txt):
http://www.fossil-scm.org/index.html/info/bc8d368b66
And as you can see here:
http://www.fossil-scm.org/index.html/timeline?d=bc8d368b&n=126
Version 1.19 is the first version to include it. So, by using the last released version prior to 1.19 (which is 1.18) and patching that the problem goes away.
I don't think I ever investigated the reason that change introduces the problem. I think I most likely ran a bisect and probably used the sqlite_docsrc repository to test with as it has the problem too and I think it's the smallest of the three (fossil, sqlite, sqlite_docsrc).
I see there's a patch to add an incremental export shortly after the problem commit. Makes me wonder if some of those patches could be cherry-picked onto the version that works. :)
Again though, if you use an incremental export and Richard shuns again you can end up with a frankenrepo that no one else can ever reproduce.