On Wed, 26 Sep 2018 18:18:23 +0200 leoutat...@gmx.fr said: > On 9/26/18 3:30 PM, Carsten Haitzler (The Rasterman) wrote: > > On Wed, 26 Sep 2018 13:37:42 +0200 leoutat...@gmx.fr said: > > > >> On 9/26/18 12:04 PM, Carsten Haitzler (The Rasterman) wrote: > >> > >>>>> you definitely have a leak. you want massif. so -massif option. then use > >>>>> massif-visualizer to look at the trace and you'll see it probably going > >>>>> "up and to the right". select a time point near the right/big end and > >>>>> start digging through the tree of allocations (most of the mem allocs > >>>>> are near the top). expand the trees for call traces. that there will > >>>>> begin to give an idea of what is going on. you also will have learned a > >>>>> new debugging technique :) > >>>>> > >>>> valgrind -massif log file is available here > >>>> https://phab.enlightenment.org/T7410 > >>> > >>> private paste - but we don't want the log. we want the massif output file. > >>> massif.out maybe or something (i have an alias that produces a specific > >>> one for me). even then i'm not sure i can usefully open it on a different > >>> system. i have never tried. thus massif-visualizer. > >>> > >>> > >> My apologize, access paste is public, now. Of course, it's massif output > >> file, 'massif.out', you can get it. > >> 'e-valgrind-massif.log' file has never been created, it's mistake. > >> Massif created a massif.out automatically. > > > > It's still private. :( > > > It's displayed public on my account, I don't understand. > I copied past raw text on phab site. I hope it's ok...
hmmm... that leak isn't in normal malloc. paths. it seems pretty ok - climbs a bit over time. this smells like something to do with drm/gbm buffers. something not being released? -- ------------- Codito, ergo sum - "I code, therefore I am" -------------- Carsten Haitzler - ras...@rasterman.com _______________________________________________ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
