In my redhat 6.0 intel with gtk 1.2.3 top reports 4 processes of 129Mb each in Aug-5 and Aug-4 nightly snapshot.

phillip, can you reproduce this?

yes, it occurs on my linux box and the one in the browser lab, except mine only spawns 3 processes, not 4. (rh52, gtk 1.2.1, will upgrade to 6.0 soon)

shaver's the top dog all over this.

Please try again with a more recent nightly (tonight's should be especially pleasant), because that 129M number is just horrible. What are you doing that causes such immense memory growth? What are your process sizes at startup? (Once the start page has rendered, I mean.)

I'm not seeing this, and we've fixed some leaks with growth patterns like this one, so I'm marking FIXED.

Mike Shaver wrote: > > I'm marking FIXED. > Otto Solares wrote: > > Don't think so... > here's a naive question: i'm not sure how gtop handles reporting user level threads, but i get the hunch that, even though there appears to be 4 processes, they all have the same exact size and therefore use shared memory. i want to believe that in the following example only 20 MB total is being used, and it is reported 4 times because there are 4 threads. am i wrong? USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND phillip 2422 0.6 12.3 20444 15868 pts/3 S 10:51 0:04 ./apprunner phillip 2423 0.0 12.3 20444 15868 pts/3 S 10:51 0:00 ./apprunner phillip 2424 0.0 12.3 20444 15868 pts/3 S 10:51 0:00 ./apprunner phillip 2425 0.0 12.3 20444 15868 pts/3 S 10:51 0:00 ./apprunner here is something for you to try, otto. when i type free, this is what i get: total used free shared buffers cached Mem: 128092 124732 3360 82808 4940 59188 -/+ buffers/cache: 60604 67488 Swap: 136040 1088 134952 so then i typed `ps aux` to get a list of every running process, piped it through cut, and made the ugliest little script (could be better in perl, but it works.) this prints out a long sum of the numbers from the fifth column output by ps: (ps aux|grep -v "USER"|cut -c 25-31|sed -e 's,$,+,' -e 's,$, \\,' && echo '0') you'll notice a lot of repeated numbers. i propose that these represent multiple references to the same block of memory used by threads. so running that command, but piping through bc, the fun calculator, i get this output: 316340, which means that either free lied to me and twice as much memory is being used by all running processes, or that, even though top and ps report what seems to be several processes (like httpd, automount, and apprunner) each using TONS of ram, they are only threads sharing the same memory space that are reported multiple times (once per thread). the net result is that what you are probably seeing is several threads of the same instance of apprunner, taking up a total of 21 MB of ram. phew, sorry for the verbosity... mike does this make sense? (or am i totally wrong?)

Where did Otto write that? Your general analysis is correct, though I think we're probably using < 21MB; a large part of that is the private /dev/zero mapping for the page manager, I believe, which doesn't take up virtual memory until we dirty the pages.

QA Contact massive update.