On Fri, Aug 04, 2000 at 08:54:05PM +0100, Lee Elliott wrote: > Hello list, > > I noticed a curious thing this evening - when I load the fetchmail man > page in the Gnome help browser, it grabs all my memory - 256MB and ramps > up swap usage until all that's gone too - another 256MB - previously > none was used. CPU utilisation runs at about 75% on both cpus (SMP > system) while this happens. > > This all takes about 10 seconds on this system. It then frees all the > memory, effectively flushing the cache and buffer memory, and displays > the fetchmail man page ok. > > Apart from the cpus being tied up, there's no other obvious effects or > consequences - nothing crashes (tried a kernel compile and running > Netscape) and the system then seems fine, although swap usage doesn't > return to 0MB immediately but seems to drop off over time, probably as > the system moves stuff back into main memory. I just logged out and > back in and that cleared most of the swap - only 9.5MB used now, down > from 64MB after loading the man page several times, dropping to 29MB > just before I logged out. > > There don't seem to be any spurious processes left hanging around. > > This happens on consecutive loads on this man page: start > Gnome help, select Man Pages, select User Commands, select fetchmail - > memory used. Use the Back button, select User Commands (again), select > fetchmail - memory used. > > I've got three discrete 2.2.17 Potatos on this system, all with the same > s/w installed - I've tried it on two of them with identical results - > reboots make no difference. At least it shows I'm keeping them in > step;) > > This doesn't happen if the fetchmail man page is loaded from a Gnome > term: 'man fetchmail'
Yep! Same on my machine ( potato, kernel 2.2.13, 64 MB + 150 MB swap ) > > dmesg gives: > > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for gnome-man2html... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for gnome-man2html... > VM: do_try_to_free_pages failed for procmeter3... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for gnome-man2html... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for init... > VM: do_try_to_free_pages failed for sh... > VM: do_try_to_free_pages failed for sh... > VM: do_try_to_free_pages failed for sh... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for sh... > VM: do_try_to_free_pages failed for procmeter3... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for tasklist_applet... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for kswapd... > VM: do_try_to_free_pages failed for gnome-man2html... > VM: do_try_to_free_pages failed for gnome-man2html... > VM: do_try_to_free_pages failed for sh... > VM: do_try_to_free_pages failed for sh... > VM: do_try_to_free_pages failed for XF86_SVGA... > VM: do_try_to_free_pages failed for sh... > VM: do_try_to_free_pages failed for sh... > VM: do_try_to_free_pages failed for procmeter3... > VM: do_try_to_free_pages failed for mount... > VM: do_try_to_free_pages failed for procmeter3... > No. Nothing like that on my machine . > I'm getting really crappy conections at the moment so I've not been able > to check the bug list, but if any one else gets the same behaviour I'll > check it out and raise a bug if neccessary. I noticed that the man page of fetchmail is quite long. Gnome Helper calls 'gnome-man2html' to convert man pages into html pages, and probably this is the program which sucks so much memory and CPU. Dunno if this counts as a bug, though. > > LeeE > -- Ciao ---------- FB
