Re: Another checksum mismatch

Philip Martin <philip@codematters.co.uk> writes:

> I have so far failed to produce a test case that results in a corrupt
> checksum, I'm tempted to build a circa 4400 client to see if I can
> reproduce it with that.

Using an r4400 client, a commit of a file that has an svn:executable
property change, but no text change causes the following valgrind
warning

$ svn ps svn:executable on wc/bar
property `svn:executable' set on 'wc/bar'
$ valgrind -q subversion/clients/cmdline/.libs/lt-svn ci wc
Sending wc/bar
==1026== Use of uninitialised value of size 4
==1026== at 0x402F0580: svn_md5_digest_to_cstring (../svn/subversion/libsvn_subr/md5.c:40)
==1026== by 0x4026A0CA: svn_wc_process_committed (../svn/subversion/libsvn_wc/adm_ops.c:234)
==1026== by 0x4025089C: svn_client_commit (../svn/subversion/libsvn_client/commit.c:1005)
==1026== by 0x804B5D2: svn_cl__commit (../svn/subversion/clients/cmdline/commit-cmd.c:100)

That looks like it would be the cause of an incorrect checksum. This
problem does not occur with r5381, I suspect it was fixed by either
r4405 or r4682.

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org