On Sat, 24 Apr 2010, Török Edwin wrote:
Does Mimedefang run clamscan under ulimit? (or is mimedefang itself constrained by some ulimits?)
I doubled the MX_MAX_RSS and MX_MAX_AS arguments in the startup script, and it seems to have taken care of the problem (which I was able to recreate, see below).
If 0.96 uses more memory than 0.95.3, then this would explain why the problem only occurred when upgrading to the new clam version.
...'make test' didn't have much to say... It is called 'make check'...
Oops. Sorry about that. Here's the output of the --disable-llvm version: make check-TESTS PASS: check_clamav PASS: check_freshclam.sh PASS: check_sigtool.sh SKIP: check_unit_vg.sh PASS: check1_clamscan.sh PASS: check2_clamd.sh PASS: check3_clamd.sh PASS: check4_clamd.sh SKIP: check5_clamd_vg.sh SKIP: check6_clamd_vg.sh SKIP: check7_clamd_hg.sh SKIP: check8_clamd_hg.sh ====================== All 7 tests passed (5 tests were not run)So, does this mean I'm missing out on some functionality, or just that the work will be done less efficiently?
However, it's not quiiiiite a fair comparison, since the Sanesecurity signatures have been pared down since my first test. I'm going to add a couple more databases and see if I can get anything to break.
I did this and it did break. Changing the mimedefang values I mentioned above unbroke it.
So, ultimately, it looks like there were two problems. One was that mimedefang was memory-starved and the other was that making with --enable-llvm caused a whole lot of error messages, even when things seemed to be working otherwise.
Both problems appear to be solved now. Thanks for all the help. Jeffrey Moskot System Administrator j...@math.miami.edu
_______________________________________________ Help us build a comprehensive ClamAV guide: visit http://wiki.clamav.net http://www.clamav.net/support/ml
