On 1/27/07, Jean-Marc Pigeon <[EMAIL PROTECTED]> wrote:
Bonjour a Tous,
I would like to know if more data has been collected on that
subject.
Regarding the test of the OP:
I made a test on x86_64, and clamav thought it detected an html file
inside the random file, a 500M+ html file, processing that takes lot
of memory/space.
0.90rc2 is smarter about that, so you might want to try using
0.90rc2/cvs when scanning /dev/urandom.
According "ps axl", X86_64 version is requesting
more than 2 times VSZ memory on X86_64 than I32 (at
starting time)... (meaningful)???
Pointers are 8 bytes X864_64 vs. 4 bytes IA32. Still it shouldn't be
2x mem usage.
The memory leak is showing up via the "free" command
where the used memory is steadily increasing over USAGE time.
Can you try if this is still an issue with 0.90rc2/cvs? [eg: on a
test-machine, sending it to scan
the same files that IA32 machine scans].
Strangely enough "ps axl" is not reporting memory
pmap is more accurate than ps on mem usage.
I have appended the output for file2 and file3 to this email.
In both cases, the increased memory consumption starts when the
following lines are printed:
The attachments didn't get through the mailing list. Could you upload
them somewhere?
LibClamAV debug: Calculated MD5 checksum: a07774f93dc2c5da62ddf502692a208e
LibClamAV debug: in cli_scanhtml()
LibClamAV debug: mmap'ed file
Seems like it tries to scan an html, which is probably the rest of the
file (a few Gigabytes probably?).
Best regards,
Edwin
_______________________________________________
Help us build a comprehensive ClamAV guide: visit http://wiki.clamav.net
http://lurker.clamav.net/list/clamav-users.html
