For vxworks, we need to find sys_mempart_id, which I haven't finished doing for all cameras. Baring typos, I think to should be safe to turn on for all that have been found.
From a quick look, sys_mempart_id could be found by looking at an eventproc named "memShow" which is a wrapper for memPartShow. The latter needs sys_mempart_id as its first param.
Done. I have also verified the already found values (A, G series) except for a570 101a (that dump seems to be missing from the "collection"). Patch attached, enables CAM_FIRMWARE_MEMINFO for all VxWorks ports which support it.
I chose not to commit this without consent, as it has impact on many ports and the only one I can test is the Ixus65.
Perhaps add the stubs_min.S values with a comment that it is untested and to enable the option in platform_camera.h to test.
Add the values in platform_camera.h commented out with instructions to enable & test.
That way the info is not lost; but won't break any existing ports.
Phil.
I don't think there are many people out there who compile CHDK from source. Probably even less who look into the source before. And probably only a fraction of those will ever mess with an old (= VxWorks) camera.
My main concern about enabling CAM_FIRMWARE_MEMINFO for all Vx ports is that I have not verified the memPartInfoGet function for most of them. That said, the sigfinder appears to be 100% certain about its address in all cases.
Memory information is used (I may have missed something):
- when someone clicks on that menu entry
- from Lua (get_meminfo)
- in PTP, performance is horrible without it