[Xfce-bugs] [XFCE 0000264]: xffm shows all files 524288GB

[xfce-bugs at xfce.org]

The following bug has been CLOSED
=======================================================================
http://bugs.xfce.org/view_bug_page.php?f_id=0000264
=======================================================================
Reporter:                   h2k
Handler:                    edscott
=======================================================================
Project:                    XFCE
Bug ID:                     0000264
Category:                   xffm
Reproducibility:            always
Severity:                   major
Priority:                   normal
Status:                     closed
=======================================================================
Date Submitted:             2004-07-21 00:05 UTC
Last Modified:              2004-09-08 00:07 UTC
=======================================================================
Summary:                    xffm shows all files 524288GB
Description: 
xffm shows all files 524288GB, nothing more to add
=======================================================================

-----------------------------------------------------------------------
 edscott - 2004-07-21 02:17 UTC 
-----------------------------------------------------------------------
I don't understand what you are talking about. Please upload a screenshot
where it shows the problem (if applicable).

-----------------------------------------------------------------------
 h2k - 2004-07-21 18:14 UTC 
-----------------------------------------------------------------------
i meant that all file sizes are 524288GB, take a look at the screenshot:
http://h2k.autonet.ee/scrsht/xffm.jpg

btw, i'm using reiserfs

-----------------------------------------------------------------------
 edscott - 2004-07-21 21:53 UTC 
-----------------------------------------------------------------------
There might have been some confusion between the off_t type and the output
in %llu format. I cleaned up the code so that no confusion should entail.
Please verify current CVS behaviour. (If problem persists, attach a copy
of your xffm/config.h file for further diagnostics).

-----------------------------------------------------------------------
 h2k - 2004-07-24 16:04 UTC 
-----------------------------------------------------------------------
i just (24.07) compiled xffm from CVS but the problem still persists

-----------------------------------------------------------------------
 edscott - 2004-07-24 19:42 UTC 
-----------------------------------------------------------------------
Could you please attach a copy of xffm/config.h to see what your computer
makes of the st_mode structure and other stuff?  Thanks.

-----------------------------------------------------------------------
 h2k - 2004-07-25 02:30 UTC 
-----------------------------------------------------------------------
sorry i couldn't upload it earlier - i was already in hurry posting last
message :)

-----------------------------------------------------------------------
 edscott - 2004-07-29 13:25 UTC 
-----------------------------------------------------------------------
The uploaded config file is practically identical to mine. Might be some
test is missing. What is the architecture on the box that exhibits the
trouble?

-----------------------------------------------------------------------
 h3k - 2004-07-29 21:51 UTC 
-----------------------------------------------------------------------
I compiled it like this: 
export CFLAGS="-O3 -ffast-math -pipe -malign-double -march=pentium2
-fomit-frame-pointer -fno-exceptions"

maby it's because this optimization?

-----------------------------------------------------------------------
 kelnos - 2004-07-29 21:58 UTC 
-----------------------------------------------------------------------
that's possible.  from the gcc man page about -ffast-math:

This option should never be turned on by any -O option since it can
result in incorrect output for programs which depend on an exact
implementation of IEEE or ISO rules/specifications for math functions.

i don't know if xffm relies on this kind of stuff, but i really wish
gentoo people wouldn't bother reporting bugs until they ditch stupid
optimisations...  try it with '-O2 -march=i686 -pipe' and report back.

-----------------------------------------------------------------------
 edscott - 2004-07-29 23:04 UTC 
-----------------------------------------------------------------------
Ah, yes. The optimization flags are the culprit. I've compiled with the
same optimizations and get 16384GB for all files. This bug will be fixed
shortly. Thanks for the extra information.

-----------------------------------------------------------------------
 edscott - 2004-07-29 23:49 UTC 
-----------------------------------------------------------------------
AFAICT, the problem is with -malign-double. When you use it as a CFLAG, the
"struct stat" returned from stat() is borked. You might want to file a bug
with the gcc guys. The other flags you use do not seem to cause any harm,
at least on my box.

-----------------------------------------------------------------------
 kelnos - 2004-09-08 00:07 UTC 
-----------------------------------------------------------------------
mass closing some old resolved bugs; please reopen if in error