It all started with a simple problem: I set the setting to dim the screen to 10s while on AC power and didn't see the screen dim. After some tracing, I saw this in src/xfpm-brightness-hal.c, lines 387-410: * after calling xfpm_brightness_get_user_timeouts, on_battery retrieves the value properly -- but on_ac returns 0 all the time. * adding a new variable into the definition (line 388: guint16 on_ac, on_battery;) and using it instead of on_ac fixes the problem * reordering the variable definitions as "guint16 on_battery, on_ac;" also fixes the problem I don't have much experience with C, but I suspect some messy business is going on with memory misalignment or stack overflows. I've compiled PM with -g3 -O0 and ran it with valgrind. Here's some output: <pre> ==1777== 168 (64 direct, 104 indirect) bytes in 2 blocks are definitely lost in loss record 100 of 215 ==1777== at 0x4006F3D: malloc (vg_replace_malloc.c:207) ==1777== by 0x767E06: XSyncListSystemCounters (in /usr/lib/libXext.so.6.4.0) ==1777== by 0x805CC78: xfpm_idle_init (xfpm-idle.c:266) ==1777== by 0x799C6E: g_type_create_instance (in /lib/libgobject-2.0.so.0.2000.4) ==1777== by 0x77E9F7: (within /lib/libgobject-2.0.so.0.2000.4) ==1777== by 0x77FBDB: g_object_newv (in /lib/libgobject-2.0.so.0.2000.4) ==1777== by 0x7807D9: g_object_new_valist (in /lib/libgobject-2.0.so.0.2000.4) ==1777== by 0x78095D: g_object_new (in /lib/libgobject-2.0.so.0.2000.4) ==1777== by 0x805CE61: xfpm_idle_new (xfpm-idle.c:321) ==1777== by 0x804FC18: xfpm_engine_init (xfpm-engine.c:572) ==1777== by 0x799C6E: g_type_create_instance (in /lib/libgobject-2.0.so.0.2000.4) ==1777== by 0x77E9F7: (within /lib/libgobject-2.0.so.0.2000.4) </pre> This one is caused by not freeing the array returned by XSyncListSystemCounters (xfpm-idle.c:266). The doc is pretty clear on that :) To fix, just insert "XSyncFreeSystemCounterList(counters);" in line 273 Unfortunately it doesn't fix my problem.. but I'll continue looking for a little while.
Sorry for the late reply, i was on vacation. Changing the variable declaration from guint16 to guint fixes the problem also, i dunno why, but anyway i have to include this fix for the next release which is coming soon.
Don't forget XSyncFreeSystemCounterList :)
(In reply to comment #2) > Don't forget XSyncFreeSystemCounterList :) I had this before in the code, but i remove the free stuff since it was causing a crash, i will retest it.
(In reply to comment #3) What were the conditions of crash? Think I could replicate it?
(In reply to comment #4) > (In reply to comment #3) > > What were the conditions of crash? Think I could replicate it? Seems that i'm not able to reproduce the problem that i had anymore, XSyncFreeSystemCounterList(counters); call added to git. Thanks.
Looks like original bug is now gone, even if I restore the code in xfpm-engine.c to the "buggy" one. I'll investigate it some more with rev 0458addb7891e79112e21bf0d09f97c627c63265, because I don't like ghost-in-the-machine bugs (implication here is that the bug just shifted elsewhere) I'll post again if I find something.
Recent version (0.8.5) doesn't seem to have any of these problems, I think this bug can be closed.