Web lists-archives.com

Bug with corruption on clone/fsck/... with large packs + 64-bit Windows, problem with usage of "long" datatype for sizes/offsets?




Hi,

on Windows 64-bit, for a repository having a .pack file > 4GB I get during cloning:

$ git clone file:///repositories/test.git test
Cloning into 'test'...
remote: Counting objects: 210294, done.
remote: error: bad object header
remote: error: bad object header
remote: fatal: packed object cae50116ebe36de8bded4811bd262d90670bde2f (stored in ./objects/pack/pack-30ba8b19b029cc6554853ae07729aa807995ebb8.pack) is corrupt
error: git upload-pack: git-pack-objects died with error.
fatal: git upload-pack: aborting due to possible repository remote: aborting due to possible repository corruption on the remote side.
corruption on the remote side.
fatal: early EOF
fatal: index-pack failed

We use git 2.14.1, the 64-bit version.

On systems which have (long int) == 64-bit datatype, all works well, e.g. Linux 64-bit.
(same goes for fscks/...., they work on this git on Linux 64-bit, not on Windows 64-bit)

It seems the Windows port did take care of using the right 64-bit offsets and all
in the OS specific parts, but things like pack-objects.c in the generic parts use "unsigned long"
like it behaves like size_t (at least on the first glance).

As I think the issue is in the generic part, I report this here, not on the Windows list,
thought it seems to be an IL32P64 related issue.

Greetings
Christoph

-- 
----------------------------- Dr.-Ing. Christoph Cullmann ---------
AbsInt Angewandte Informatik GmbH      Email: cullmann@xxxxxxxxxx
Science Park 1                         Tel:   +49-681-38360-22
66123 Saarbrücken                      Fax:   +49-681-38360-20
GERMANY                                WWW:   http://www.AbsInt.com
--------------------------------------------------------------------
Geschäftsführung: Dr.-Ing. Christian Ferdinand
Eingetragen im Handelsregister des Amtsgerichts Saarbrücken, HRB 11234