| situ | hmm I am seeing tearing if I run surfaceflinger. | 04:20 |
|---|---|---|
| situ | This probably means it's not a problem with our qpa plugin. | 04:20 |
| *** tbr has quit IRC | 06:11 | |
| situ | Not able to debug with gdb :( | 14:57 |
| MSameer | situ: what are you trying to debug? | 15:35 |
| situ | MSameer: Has 'info locals' ever worked for you ? | 15:35 |
| MSameer | situ: are you trying to debug a bionic binary? | 15:36 |
| situ | MSameer: No, qpa plugin | 15:36 |
| MSameer | what's the problem then? | 15:37 |
| MSameer | no symbols? | 15:37 |
| situ | Yeah, it does not show anything when I run 'info locals' | 15:39 |
| MSameer | perhaps they are optimized out | 15:39 |
| MSameer | you need the -debuginfo package too | 15:39 |
| situ | I did install them | 15:40 |
| situ | Both debugsource and debuginfo packages. | 15:40 |
| situ | morphis: Can you poinit me to your commits which fixed compositor on webOS ? | 19:44 |
| morphis | situ: on the android side? | 19:45 |
| situ | morphis: Yes | 19:46 |
| situ | I am afraid this line is causing the crash | 19:46 |
| situ | https://github.com/CyanogenMod/android_hardware_qcom_display/blob/cm-10.2/libhwcomposer/hwc.cpp#L89 | 19:46 |
| morphis | change it to use numDisplays | 19:48 |
| morphis | https://github.com/webOS-ports/android_hardware_qcom_display-caf/commit/c02650576fdeed8af1f74d4dc49f07b8a8c4f577 | 19:48 |
| situ | morphis: Yeah, I just want to see what all changes you have made to fix compositor for webOS | 19:48 |
| morphis | situ: second one on android side is https://github.com/webOS-ports/android_hardware_qcom_display-caf/commit/ec7b601a9debef88d06cd8684801baef4597ad3e | 19:49 |
| morphis | but I doubt you need it | 19:49 |
| morphis | it's dirty as hell but works to avoid the GLES binaries allocating more memory as they should (it happens so far only in one specific case) | 19:49 |
| situ | morphis: You are right. At this point I am just concerned about the crash which I see when I tweak value of numDisplays. | 19:49 |
| morphis | situ: everything else is http://bpaste.net/show/8qA5QtUAFrAKEeBq8gDW/ | 19:50 |
| situ | I am not sure if I need changes in hwcomposer. | 19:51 |
| situ | lbt_: Now I am finding that it's better that we clone all the repos under build tree even if we don't need them. So if we need to change something outside our current changes, it's easier to resolve dependencies. | 19:56 |
| situ | lbt_: That said, it would be better if we revert our changes from this commit https://github.com/mer-hybris/android_build/commit/ac574b74890d0949d626dc4d0c96f143fa60ae9f | 19:58 |
| situ | morphis: Strange code https://github.com/CyanogenMod/android_hardware_qcom_display/blob/cm-11.0/msm8974/libhwcomposer/hwc.cpp#L266 | 20:00 |
| situ | They are iterating in reverse order. | 20:00 |
| situ | morphis: I see these messages in logcat output before crash http://pastie.org/9359114 | 20:24 |
| morphis | situ: you tried to run lipstick with gdb and generate a backtrace? | 20:25 |
| situ | morphis: I never get a usable backtrace. | 20:25 |
| morphis | you need to load the android symbols files at the right place | 20:26 |
| morphis | they are not loaded automatically | 20:26 |
| morphis | use something like http://bpaste.net/show/AFamECJZVN9fvugNhGMZ/ to generate the add commands to load the android symbol files from /system/symbols | 20:26 |
| situ | morphis: will try that later, it looks like 'MAX_DISPLAYS' is blindly being used at several more places. | 20:30 |
| lbt_ | situ: I'm a bit afk at the moment- hurt my back | 20:53 |
| situ | lbt_: Enjoy the match ;) | 20:54 |
| lbt_ | I'd really rather not revert that change - the point is to clone and manage a minimal subset of android. Many of our users don't have huge amounts of space or the bandwidth | 20:55 |
| lbt_ | hehe - nah, don't do football | 20:55 |
| lbt_ | luckily I can still stand and hence I can still dance so Tango is OK | 20:55 |
| situ | lbt_: But it will help in debugging new packages. | 20:55 |
| situ | I agree we can remove all the application repositories. | 20:55 |
| lbt_ | I have no problem growing that list if we need ot | 20:56 |
| situ | But at least we should keep all the lower layers. | 20:56 |
| lbt_ | why ? | 20:56 |
| situ | lbt_: So, right now I am debugging hwcomposer and I had to tweak build/core/main.mk file to resolve dependencies. | 20:56 |
| lbt_ | yeah | 20:58 |
| lbt_ | we're competent to do that | 20:58 |
| lbt_ | the goal is to have a tree which is minimal to 'build' | 20:58 |
| lbt_ | and I agree the more advanced of our team should document how to extend the package sert | 20:58 |
| lbt_ | set | 20:58 |
| situ | hmm ok. | 20:59 |
| lbt_ | to make it less of an adventure to debug :) | 20:59 |
| * lbt_ is still chasing segfaults in init which are down to 4.4 changes | 20:59 | |
| lbt_ | but I mainly know what's going on now | 20:59 |
| situ | lbt_: Is 4.4 changes very different from what we have done for cm11 ? | 21:02 |
| lbt_ | I haven't fully compared them - there was not much needed really | 21:04 |
| lbt_ | I was looking at how to do manifest management | 21:04 |
| lbt_ | it took me a while because I've had a lot of infra stuff to deal with | 21:04 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!