[Bug 8666] --debug=all9 fail

[samba-bugs at samba.org]

https://bugzilla.samba.org/show_bug.cgi?id=8666

--- Comment #1 from Chris Dunlop <chris at onthe.net.au> 2011-12-16 13:40:01 UTC ---
(In reply to comment #0)
> There are at least 2 recursive calls when running --debug=all9:
> 
> (gdb) bt
> #1  0x0000000000434996 in perform_io (needed=60, flags=4) at ../rsync/io.c:608
> #2  0x0000000000435b71 in send_msg (code=MSG_INFO, buf=0x7fffa60c7da0
> "[receiver] perform_io(60, msgroom) needs to flush 33\n", len=53, convert=0) at
> ../rsync/io.c:955
> #3  0x0000000000429840 in rwrite (code=FINFO, buf=0x7fffa60c7da0 "[receiver]
> perform_io(60, msgroom) needs to flush 33\n", len=53, is_utf8=0) at
> ../rsync/log.c:277
> #4  0x0000000000429f37 in rprintf (code=FINFO, format=0x46c8c8 "[%s]
> perform_io(%ld, msgroom) needs to flush %ld\n") at ../rsync/log.c:433
> #5  0x0000000000434996 in perform_io (needed=60, flags=4) at ../rsync/io.c:608
> 
> ...and so on ad-infinitum. If the rprintf() at io.c:608 is patched out, you get
> this instead:
> 
> (gdb) bt
> #0  0x0000000000429a11 in rwrite (code=FINFO, buf=0x7fff3006b8e0 "[generator]
> reduced size of iobuf.msg (-3)\n", len=43, is_utf8=0) at ../rsync/log.c:321
> #1  0x0000000000429f37 in rprintf (code=FINFO, format=0x46c6e8 "[%s] reduced
> size of %s (-%d)\n") at ../rsync/log.c:433
> #2  0x00000000004345a5 in reduce_iobuf_size (out=0x6839c0, new_size=32765) at
> ../rsync/io.c:488
> #3  0x0000000000435b46 in send_msg (code=MSG_INFO, buf=0x7fff3006d770
> "[generator] reduced size of iobuf.msg (-3)\n", len=43, convert=0) at
> ../rsync/io.c:964
> #4  0x0000000000429a16 in rwrite (code=FINFO, buf=0x7fff3006d770 "[generator]
> reduced size of iobuf.msg (-3)\n", len=43, is_utf8=0) at ../rsync/log.c:321
> 
> ...and so on ad-infinitum. If the rprintf() at io.c:488 is patched out you no
> longer get these recursive debug calls (but might there be others lurking?).
>
> For what it's worth, this was whilst running:
> 
> /tmp/rsync-testing --rsync-path=/tmp/rsync-testing -aHXX --numeric-ids
> --debug=all9 --link-dest=../A --link-dest=../B C/ server:$PWD/C/
> 
> ...where the directory hierarchies are quite substantial.

After those calls are patched out but still with --debug=all9, I get a 29K line
debug output that ends in:

unexpected tag 95 [sender/inc]
[sender] _exit_cleanup(code=12, file=../rsync/io.c, line=1615): entered
rsync error: error in rsync protocol data stream (code 12) at
../rsync/io.c(1615) [sender=3.1.0dev]
[sender] _exit_cleanup(code=12, file=../rsync/io.c, line=1615): about to call
exit(12)

...and the receiving side got only as far as creating the top level directory.
Same
thing down to --debug=all4. Even down at --debug=all1 I get an aborted
transfer:

[sender] send_msg(0, 65532)
rsync: [sender] write error: Broken pipe (32)
[sender] got msg=2, len=31
[generator] send_msg(0, 22907)
rsync error: error in rsync protocol data stream (code 12) at
../rsync/io.c(823) [sender=3.1.0dev]
[sender] _exit_cleanup(code=10, file=../rsync/io.c, line=823): about to call
exit(12)

Looks like something in --debug=all is writing into the io stream?  Without
--debug=all the transfer progresses fine.


But that doesn't help me debug my problems with the new --link-by-hash stuff
(all previous tests were on unpatched master).  At least I now know most of the
problems I thought I was seeing with --link-by-hash were actually because of
--debug=all!  Oh well, I guess I'll have to start using explicit debug flags.

Cheers,

Chris

-- 
Configure bugmail: https://bugzilla.samba.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.