Re: Odd decision of git-pasky-0.7 to do a merge

On Fri, Apr 29, 2005 at 08:36:16AM +0100, Russell King wrote:

This is what happened this morning to a tree which only gets pulled
from Linus' tree.  No other changes are ever committed to it.

Why it decided that a merge was necessary is beyond me.  Any ideas?
Did Linus forget to merge his tree properly?

Ok, there's something weird going on here.  On hera:

rmk@hera:/pub/scm/linux/kernel/git/torvalds/linux-2.6.git(0)$ SHA1_FILE_DIRECTORY=./objects fsck-cache
expect dangling commits - potential heads - due to lack of head information
dangling commit c60c390620e0abb60d4ae8c43583714bda27763f
rmk@hera:/pub/scm/linux/kernel/git/torvalds/linux-2.6.git(0)$ md5sum objects/03/97236d43e48e821cce5bbe6a80a1a56bb7cc3a
fd1369dbbd494b1839ad3d633a5d088e  objects/03/97236d43e48e821cce5bbe6a80a1a56bb7cc3a
rmk@hera:/pub/scm/linux/kernel/git/torvalds/linux-2.6.git(0)$ md5sum objects/9e/734775f7c22d2f89943ad6c745571f1930105f
b3b788308b7f137231652b5e97cb08c0  objects/9e/734775f7c22d2f89943ad6c745571f1930105f

In my repository:

$ fsck-cache
bad sha1 file: .git/objects/03/97236d43e48e821cce5bbe6a80a1a56bb7cc3a
root 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2
bad sha1 file: .git/objects/9e/734775f7c22d2f89943ad6c745571f1930105f
expect dangling commits - potential heads - due to lack of head information
dangling commit 7598082671ce811737eca8119968dc056d78ac4b
$ md5sum objects/03/97236d43e48e821cce5bbe6a80a1a56bb7cc3a
fd1369dbbd494b1839ad3d633a5d088e  objects/03/97236d43e48e821cce5bbe6a80a1a56bb7cc3a
$ md5sum objects/9e/734775f7c22d2f89943ad6c745571f1930105f
b3b788308b7f137231652b5e97cb08c0  objects/9e/734775f7c22d2f89943ad6c745571f1930105f

These two objects appear to be tags according to cat-file -t on hera,
so these probably aren't the cause - it's more likely that git-pasky-0.7
doesn't know how to handle these objects.

Tracked branch, applying changes...
Merging e8108c98dd6d65613fa0ec9d2300f89c48d554bf -> c60c390620e0abb60d4ae8c43583714bda27763f
	to bdceb6a0162274934386f19f3ea5a9d44feb0b20...

Ok.  So.

rmk@hera:/pub/scm/linux/kernel/git/torvalds/linux-2.6.git(0)$ SHA1_FILE_DIRECTORY=./objects merge-base c60c390620e0abb60d4ae8c43583714bda27763f bdceb6a0162274934386f19f3ea5a9d44feb0b20
bdceb6a0162274934386f19f3ea5a9d44feb0b20

$ merge-base c60c390620e0abb60d4ae8c43583714bda27763f bdceb6a0162274934386f19f3ea5a9d44feb0b20
e8108c98dd6d65613fa0ec9d2300f89c48d554bf

Hmm.  At this point, I think we need something which graphically
shows the layout of the commit tree to work out what's going on,
and why my merge-base picked that particular commit id.

However, at this point, I think I can say that git obviously isn't
working as designed for me.

-- 
Russell King