Bug#1041647: Further gdb debugging (more #2)

2023-07-21 Thread Tj
After solving the relative path issue I now have better info: (gdb) s __vaDriverInit_1_17 (ctx=0x555702b0) at ../src/gallium/frontends/va/context.c:123 123if (!ctx) (gdb) n 126drv = CALLOC(1, sizeof(vlVaDriver)); (gdb) n 127if (!drv) (gdb) n 130switch

Bug#1041647: Further gdb debugging (more)

2023-07-21 Thread Tj
The mesa-va-drivers-dbgsym package seems to have incorrect (relative to build) paths stored which prevents gdb showing the source lines when it has the path to the mesa-22.6.3 source, however by blindly stepping and then looking at the source it narrows down the cause: (gdb) s 126 in

Bug#1041647: Further gdb debugging

2023-07-21 Thread Tj
Using export LIBVA_MESSAGING_LEVEL=2; export LIBVA_DRIVER_NAME=radeonsi; gdb --directory . --directory ../libva-2.17.0 --directory ../mesa-22.3.6 --args /usr/bin/vainfo I've stepped through vainfo and focused on va_openDriver() which eventually leads to: (gdb) n libva info: Found init