8b01247 | Daniel Fort | 27 December 2018, 16:56:09 UTC | state-object: updated https://www.magiclantern.fm/forum/index.php?topic=15895.msg209800#msg209800 | 27 December 2018, 16:56:09 UTC |
2230940 | Daniel Fort | 27 December 2018, 14:47:40 UTC | state-object: INPUT_SET_IMAGE_VRAM_PARAMETER_MUTE_FLIP_CBR no longer used https://www.magiclantern.fm/forum/index.php?topic=15895.msg209792#msg209792 | 27 December 2018, 14:47:40 UTC |
6b7db3e | Daniel Fort | 14 December 2018, 15:58:27 UTC | Change EOSM2 const.h value for DRYOS_ASSERT_HANDLER - thanks @ccritix | 14 December 2018, 15:58:27 UTC |
3351ccb | Daniel Fort | 23 February 2018, 05:55:02 UTC | Dirty pixels fix | 23 February 2018, 05:55:02 UTC |
7b4df3c | Daniel Fort | 21 February 2018, 19:50:34 UTC | Forgot to save changes to dm-spy.c and the bottom bar hack isn't needed for logging. | 21 February 2018, 19:50:34 UTC |
5f0fd59 | Daniel Fort | 21 February 2018, 19:25:53 UTC | Test as directed by a1ex: https://www.magiclantern.fm/forum/index.php?topic=15895.msg197660#msg197660 Needs CONFIG_DEBUG_INTERCEPT and CONFIG_MMIO_TRACE | 21 February 2018, 19:25:53 UTC |
770e9bd | Daniel Fort | 21 February 2018, 13:40:31 UTC | Merged EOSM2.103_io_trace into EOSM2.103_io_trace_full | 21 February 2018, 13:40:31 UTC |
94044e0 | Daniel Fort | 21 February 2018, 12:41:14 UTC | Merged in latest EOSM2.103_wip | 21 February 2018, 12:41:14 UTC |
4e6390d | Daniel Fort | 21 February 2018, 12:34:31 UTC | Merged in latest io_trace changes. | 21 February 2018, 12:34:31 UTC |
ae188c4 | Daniel Fort | 21 February 2018, 03:03:24 UTC | Revert changes made for minimal test in the previous commit. | 21 February 2018, 03:03:24 UTC |
b484653 | Daniel Fort | 21 February 2018, 03:01:58 UTC | Test on minimal: https://www.magiclantern.fm/forum/index.php?topic=15895.msg197605#msg197605 | 21 February 2018, 03:01:58 UTC |
ae4157b | Daniel Fort | 21 February 2018, 01:56:19 UTC | Reverted debugging changes. | 21 February 2018, 01:56:19 UTC |
68308a9 | Daniel Fort | 20 February 2018, 23:47:02 UTC | Formatted comments. | 20 February 2018, 23:47:02 UTC |
becca8e | Daniel Fort | 20 February 2018, 14:06:21 UTC | raw.c debugging - temporary, remember to revert this. | 20 February 2018, 14:06:21 UTC |
a786d86 | Daniel Fort | 20 February 2018, 13:30:30 UTC | PREFERRED_RAW_TYPE: This didn't work but is worth noting until the real fix is found. | 20 February 2018, 13:30:30 UTC |
fb44e17 | Daniel Fort | 16 February 2018, 00:25:00 UTC | This extends the time between starting the camera and LiveView freezing up. Still need to look into it some more. | 16 February 2018, 00:25:00 UTC |
817b0a1 | alex@thinkpad | 15 February 2018, 10:40:13 UTC | dm-spy: option to properly timestamp "sparse" messages (useful if there may be more than one second between two messages, as the hardware timer wraps around at 1048576 us) | 15 February 2018, 10:40:13 UTC |
f86d7d4 | Daniel Fort | 15 February 2018, 01:42:32 UTC | Found value for PHOTO_CMOS_ISO_START | 15 February 2018, 01:42:32 UTC |
ab227ff | alex@thinkpad | 15 February 2018, 00:29:46 UTC | dm-spy: LED feedback while pretty-printing messages | 15 February 2018, 00:29:46 UTC |
01373ea | Daniel Fort | 14 February 2018, 05:30:01 UTC | Found FRAME_CMOS_ISO_START using adtg_gui | 14 February 2018, 05:30:01 UTC |
590750f | alex@thinkpad | 14 February 2018, 00:34:10 UTC | dm-spy-extra: use STUB_ENTRY for EDMAC functions (no more need to hardcode these stubs) | 14 February 2018, 00:34:10 UTC |
ec833a5 | alex@thinkpad | 14 February 2018, 23:39:27 UTC | dm-spy: custom logging function for SetEDmac; fine-tuned other EDMAC helpers | 14 February 2018, 23:39:27 UTC |
9f5f724 | alex@thinkpad | 14 February 2018, 23:33:59 UTC | dm-spy-extra: fine-tuned configuration for photo capture logging (5D3.113) | 14 February 2018, 23:33:59 UTC |
4aafaaf | alex@thinkpad | 14 February 2018, 00:40:53 UTC | dm-spy-extra: old mmio_log experiments no longer needed (still keeping this function for compatibility) | 14 February 2018, 00:40:53 UTC |
dec6710 | alex@thinkpad | 14 February 2018, 23:13:51 UTC | dm-spy: unwrap the timer and print timestamps as ss.mmm.uuu (s=second, m=millisecond, u=microsecond) | 14 February 2018, 23:13:51 UTC |
8cc54cf | alex@thinkpad | 14 February 2018, 22:47:18 UTC | io_trace: decode usual LDR/STR addressing modes (most of the A5.2 in ARM ARM); also store Rm during data collection (values may not be always valid; decoded later) | 14 February 2018, 22:47:18 UTC |
f89bdb2 | alex@thinkpad | 14 February 2018, 21:54:44 UTC | io_trace: dummy wrappers for use with CONFIG_MMIO_TRACE=n | 14 February 2018, 21:54:44 UTC |
bfa4b44 | alex@thinkpad | 14 February 2018, 21:34:10 UTC | dm-spy-extra: removed the "from %x" mentions, as we can now override PC for all messages | 14 February 2018, 21:34:10 UTC |
846b67f | alex@thinkpad | 14 February 2018, 20:48:17 UTC | dm-spy: fix minor regression (missing class/level values in DebugMsg's) | 14 February 2018, 20:48:17 UTC |
2f4841d | alex@thinkpad | 13 February 2018, 20:36:23 UTC | io_trace: updated comments about TCM usage | 13 February 2018, 20:36:23 UTC |
f9fb136 | alex@thinkpad | 13 February 2018, 20:03:37 UTC | dm-spy: fix incorrect address in messages from DebugMsg (broken after calling io_trace_log_flush before reading LR) | 13 February 2018, 20:03:37 UTC |
e404344 | alex@thinkpad | 13 February 2018, 17:16:05 UTC | io_trace: moved the stack in some unused area in the TCM; documented TCM usage the stack configured for Data Abort by Canon bootloader may overwrite IRQ handler table - on 5D3 we were just lucky (todo: make sure it's really unused on all models, using some automated test) | 13 February 2018, 17:16:05 UTC |
6a153e6 | alex@thinkpad | 13 February 2018, 09:56:03 UTC | io_trace: removed uncacheable jumps and clean the cache without affecting the cache hacks (patch from g3gg0; updated some comments) | 13 February 2018, 09:56:03 UTC |
512f5ed | alex@thinkpad | 13 February 2018, 00:20:36 UTC | dm-spy: hide MMIO access performed by EDMAC interrupt logger; io_trace: helpers to pause/resume MMIO logging | 13 February 2018, 00:20:36 UTC |
d55ad67 | alex@thinkpad | 12 February 2018, 11:42:45 UTC | dm_spy: re-enabled EDMAC info on interrupts | 12 February 2018, 11:42:45 UTC |
7e6dd89 | alex@thinkpad | 12 February 2018, 11:34:59 UTC | dm-spy: attempt to log addresses where interrupts were triggered (what task was interrupted and where exactly) | 12 February 2018, 11:34:59 UTC |
201e88e | alex@thinkpad | 12 February 2018, 08:56:31 UTC | io_trace: run the trapped instruction from uncacheable memory (no more need to clear the caches -> no more interference with cache hacks!) 0x000b0e74: e92d4000 push {lr} 0x000b0e78: e38ff101 orr pc, pc, #1073741824 ; 0x40000000 0x400b0e80: e5810004 str r0, [r1, #4] ; [0xC0222004] <- 0x10 0x400b0e84: e3cff101 bic pc, pc, #1073741824 ; 0x40000000 0x000b0e8c: e8bd4000 pop {lr} | 12 February 2018, 08:56:31 UTC |
4f61990 | alex@thinkpad | 11 February 2018, 23:43:53 UTC | dm-spy, io_trace: timestamp messages without logging their MMIO timer read (fixes some out-of-order timestamps and reduces unnecessary verbosity) | 11 February 2018, 23:43:53 UTC |
898cdbe | alex@thinkpad | 11 February 2018, 22:04:56 UTC | dm-spy, io_trace: major rework to reduce CPU usage, to be able to log LiveView and photo capture without issues - store messages in a binary data structure, format things like task name, timestamp and PC later - sync with other loggers that may use their own data structure (currently io_trace, todo: EDMAC) - non-sprintf wrappers (debug_logstr, debug_loghex) are now line-buffered - use shoot_malloc for output buffer (formatting messages takes a few seconds) - can log LiveView (25/30 FPS, not 50/60 FPS), can log photo capture without interference (image is still good!) - some features are broken (CONFIG_DEBUG_INTERCEPT_STARTUP with CONFIG_MMIO_LOG, CONFIG_DEBUG_INTERCEPT_STARTUP_BLINK, PRINT_STACK) | 11 February 2018, 22:04:56 UTC |
ea051b7 | alex@thinkpad | 11 February 2018, 22:04:56 UTC | NotifyBox: log messages to QEMU console | 11 February 2018, 22:04:56 UTC |
5db693c | alex@thinkpad | 11 February 2018, 13:19:10 UTC | snprintf: assert if size is negative | 11 February 2018, 13:19:10 UTC |
400bba9 | alex@thinkpad | 11 February 2018, 10:32:36 UTC | dm-spy: refactored CONFIG_DEBUG_INTERCEPT and CONFIG_DEBUG_INTERCEPT_STARTUP to use the same (common) code | 11 February 2018, 10:32:36 UTC |
beb3dce | alex@thinkpad | 11 February 2018, 23:13:11 UTC | dm-spy-extra: some options to log photo capture or LiveView with full MMIO trace (not working well; these processes are very sensitive to timing; our logging code is way too slow) | 11 February 2018, 23:13:11 UTC |
9753afd | Daniel Fort | 09 February 2018, 00:01:44 UTC | Change comment to match the changed code. | 09 February 2018, 00:01:44 UTC |
0eaf0f4 | Daniel Fort | 08 February 2018, 23:56:57 UTC | Changed from DEBUGMSG to PATCH Doesn't seem to make much difference at this stage. Camera keeps freezing up with either one so let's go with the newer method. | 08 February 2018, 23:56:57 UTC |
14a468d | alex@thinkpad | 08 February 2018, 22:47:29 UTC | io_trace: workaround to prevent crash on instructions that modify LR fixme: run the instruction in the original mode (IRQ, user, supervisor etc) http://www.magiclantern.fm/forum/index.php?topic=2388.msg196991#msg196991 | 08 February 2018, 22:47:29 UTC |
1fd24fc | alex@thinkpad | 08 February 2018, 22:47:29 UTC | io_trace: identify lost data (if index wraps around); use larger buffer todo: branch-less saturated arithmetic for buffer_index | 08 February 2018, 22:47:29 UTC |
4a1bd0e | Daniel Fort | 08 February 2018, 22:39:42 UTC | CONFIG_LVAPP_HACK_DEBUGMSG is no longer a valid option. https://www.magiclantern.fm/forum/index.php?topic=2388.msg196991#msg196991 | 08 February 2018, 22:39:42 UTC |
69580a9 | Daniel Fort | 07 February 2018, 21:42:47 UTC | Missed one in beep.c | 07 February 2018, 21:42:47 UTC |
844aabb | Daniel Fort | 07 February 2018, 21:35:28 UTC | Merged in latest io_trace changes. | 07 February 2018, 21:35:28 UTC |
8a93dda | Daniel Fort | 07 February 2018, 21:33:46 UTC | Merge in EOSM2 | 07 February 2018, 21:33:46 UTC |
30451e8 | Daniel Fort | 07 February 2018, 20:39:31 UTC | More tinkering without resolving any issues. | 07 February 2018, 20:39:31 UTC |
d9ec4b4 | alex@thinkpad | 07 February 2018, 20:25:48 UTC | io_trace: added timestamp, interrupt ID, nicer output starts to resemble the output of '-d io' from QEMU for some reason it doesn't crash any more, but I'm not sure what I did to fix it :) EDMAC and JPCORE also working, even from startup | 07 February 2018, 20:25:48 UTC |
1e6ab41 | Daniel Fort | 07 February 2018, 16:16:06 UTC | Adding changes from pull request #897 | 07 February 2018, 16:16:06 UTC |
89b5bbf | Daniel Fort | 07 February 2018, 16:09:25 UTC | Trying to track down an issue with the shutter speed not displaying properly in movie mode. | 07 February 2018, 16:09:25 UTC |
5a7a22f | Daniel Fort | 07 February 2018, 14:55:58 UTC | Another formula in a constant that might be causing issues. | 07 February 2018, 14:55:58 UTC |
57884e0 | Daniel Fort | 07 February 2018, 14:13:58 UTC | Formatting changes. Found other constants that might not work if the value has a formula in it. | 07 February 2018, 14:13:58 UTC |
5a66d64 | alex@thinkpad | 07 February 2018, 10:02:18 UTC | io_trace: helper to define memory regions without having to look up their sizes; some examples of peripherals that could be successfully logged | 07 February 2018, 10:02:18 UTC |
10d9202 | alex@thinkpad | 07 February 2018, 08:33:41 UTC | io_trace: config option for Makefile.user enable with CONFIG_MMIO_TRACE=y, to be used with either CONFIG_DEBUG_INTERCEPT=y or CONFIG_DEBUG_INTERCEPT_STARTUP=y note: EDMAC logging works with the former, range 0xC0F0001F doesn't crash (just an example) guess: LCLK must be active when starting logging? | 07 February 2018, 08:33:41 UTC |
944f75b | alex@thinkpad | 07 February 2018, 07:02:02 UTC | io_trace: interrupt hooks apparently not needed it seems to log even more registers without them e.g. C02200BC: ~100 before, ~1000 after (5D3, regular startup sequence) | 07 February 2018, 07:02:02 UTC |
cdd0cad | alex@thinkpad | 07 February 2018, 08:14:00 UTC | io_trace: simplified install/uninstall code | 07 February 2018, 08:14:00 UTC |
94e46b7 | alex@thinkpad | 07 February 2018, 07:44:54 UTC | io_trace: proof of concept working, can log GPIO and SIO many other memory regions are crashing (EDMAC, etc) | 07 February 2018, 07:44:54 UTC |
86e8fec | alex@thinkpad | 07 February 2018, 07:36:39 UTC | io_trace: custom stack probably not needed | 07 February 2018, 07:36:39 UTC |
b0cab8d | alex@thinkpad | 07 February 2018, 08:11:33 UTC | io_trace: cache sync required before reconfiguring the protection region? (it no longer crashes on camera after this change) | 07 February 2018, 08:11:33 UTC |
763b886 | Daniel Fort | 07 February 2018, 06:29:24 UTC | This shouldn't make a difference but apparently it does. LUA tests can now complete. | 07 February 2018, 06:29:24 UTC |
f0e58e5 | alex@thinkpad | 06 February 2018, 20:11:25 UTC | io_trace: refactored to use static variables/functions; removed io_trace_ prefix from local symbols to reduce noise in variable names; use uint32_t instead of unsigned int | 06 February 2018, 20:11:25 UTC |
a530a9b | alex@thinkpad | 06 February 2018, 20:04:45 UTC | io_trace: whitespace | 06 February 2018, 20:04:45 UTC |
c9e0506 | Daniel Fort | 06 February 2018, 17:52:08 UTC | Pretty formatting for hex numbers. | 06 February 2018, 17:52:08 UTC |
71a2a7c | alex@thinkpad | 06 February 2018, 16:51:54 UTC | io_trace: able to trap a few accesses in the MMIO region todo: - log the value read from the MMIO register (ideally without re-reading that register) - once some access is made, re-enable the memory protection for the next instruction (currently it's only enabled at the next interrupt, which misses most of the action) - works in QEMU, crashes on camera | 06 February 2018, 16:51:54 UTC |
f351bb6 | alex@thinkpad | 06 February 2018, 16:51:54 UTC | io_trace: started from mem_prot code, not working yet (renamed mem_prot_* to io_trace_*, removed GUI code, added hooks to dm-spy) Goal: log all MMIO register accesses (in particular, reads) alongside other debug messages | 06 February 2018, 16:51:54 UTC |
7a96400 | Daniel Fort | 05 February 2018, 16:37:21 UTC | Found SRM_BUFFER_SIZE will be useful later. | 05 February 2018, 16:37:21 UTC |
368aaa6 | Daniel Fort | 05 February 2018, 16:23:34 UTC | FRAME_SHUTTER_BLANKING_WRITE is not defined on EOSM. Check back later to see if it can be used. | 05 February 2018, 16:23:34 UTC |
ea2c332 | Daniel Fort | 05 February 2018, 07:33:59 UTC | setup to experiment with different edmac channels. | 05 February 2018, 07:33:59 UTC |
5d65eda | Daniel Fort | 05 February 2018, 02:28:30 UTC | Merged unified into wip | 05 February 2018, 02:28:30 UTC |
e9242c6 | Daniel Fort | 05 February 2018, 02:10:26 UTC | Start with EOSM values. | 05 February 2018, 02:10:26 UTC |
82a9752 | Daniel Fort | 04 February 2018, 22:52:42 UTC | Find the latest version of exiftool. | 04 February 2018, 22:52:42 UTC |
f14bd34 | alex@thinkpad | 03 February 2018, 20:55:07 UTC | 100D, 70D: enough memory for 2MB startup logs | 03 February 2018, 20:55:07 UTC |
8066448 | alex@thinkpad | 03 February 2018, 20:54:47 UTC | Merged unified into dm-spy-experiments (Makefile updates, install_qemu) | 03 February 2018, 20:54:47 UTC |
c790045 | alex@thinkpad | 02 February 2018, 21:07:41 UTC | Merged in garry23/dof-fix-for-focus.c/garry23/focusc-edited-online-with-bitbucket-by-g-1495645868417 (pull request #831) Changed the DoF equations to better handle integer math and used a different source for the equations. | 02 February 2018, 21:07:41 UTC |
ff43dd5 | alex@thinkpad | 02 February 2018, 21:07:04 UTC | Dummy merge (reference changeset for make install_qemu) | 02 February 2018, 21:07:04 UTC |
e0ceefa | alex@thinkpad | 02 February 2018, 21:05:58 UTC | Merged in makefile-updates (pull request #898) Makefile updates part 2 (minor fixes, updated make install_qemu) | 02 February 2018, 21:05:58 UTC |
7553f12 | alex@thinkpad | 02 February 2018, 20:56:20 UTC | focus.c: fix possible overflow when computing hyperfocal (corner case) examples (focal_len, focus_dist): 73 4689 90 7123 91 7282 98 8444 ... | 02 February 2018, 20:56:20 UTC |
8978943 | alex@thinkpad | 02 February 2018, 20:07:32 UTC | focus.c: fix typo pointed out by garry23 http://www.magiclantern.fm/forum/index.php?topic=9741.msg196728#msg196728 | 02 February 2018, 20:07:32 UTC |
26937f3 | alex@thinkpad | 01 February 2018, 00:18:34 UTC | dm-spy: option to save a stack trace (backtrace) for the logged functions to enable: BKT flag, e.g. { 0xFF0F8BA8, "evfComAct", 4 | BKT } | 01 February 2018, 00:18:34 UTC |
3e4beec | Daniel Fort | 30 January 2018, 20:49:35 UTC | a1ex pointed out a mistake in one of the stubs. https://www.magiclantern.fm/forum/index.php?topic=15895.msg196560#msg196560 | 30 January 2018, 20:49:35 UTC |
e70e391 | alex@thinkpad | 30 January 2018, 19:50:08 UTC | Makefile: updated rules for "make install_qemu" (for reference, should anyone ever want to use them on older changesets) Before using them, qemu has to be installed first in qemu-eos (from the qemu branch) QEMU installation directory can be changed from the QEMU_DIR environment variable, or from Makefile.user | 30 January 2018, 19:50:08 UTC |
7a7e744 | alex@thinkpad | 30 January 2018, 00:38:48 UTC | Makefile: moved default QEMU directory to qemu-eos | 30 January 2018, 00:38:48 UTC |
11bfc4a | alex@thinkpad | 30 January 2018, 19:50:08 UTC | Makefile: also updated QEMU_INSTALLED_DBI_DIR to use QEMU_PATH | 30 January 2018, 19:50:08 UTC |
1b9f438 | alex@thinkpad | 30 January 2018, 19:34:26 UTC | Makefile: fix copy/paste error | 30 January 2018, 19:34:26 UTC |
753f229 | alex@thinkpad | 30 January 2018, 18:39:07 UTC | Makefile: fix cleaning up after "make zip"; removed warning | 30 January 2018, 18:39:07 UTC |
12b4571 | alex@thinkpad | 30 January 2018, 00:38:48 UTC | Makefile: allow changing QEMU installation directory from Makefile.user and/or the environment (QEMU_DIR) | 30 January 2018, 00:38:48 UTC |
481be18 | alex@thinkpad | 29 January 2018, 22:03:42 UTC | Makefile: zip and install_qemu targets for modules | 29 January 2018, 22:03:42 UTC |
cac52db | alex@thinkpad | 30 January 2018, 18:19:37 UTC | Makefile: always clean after running install_check http://www.magiclantern.fm/forum/index.php?topic=991.msg194444#msg194444 | 30 January 2018, 18:19:37 UTC |
c7d9565 | alex@thinkpad | 30 January 2018, 18:08:52 UTC | Makefile: always rebuild version.o | 30 January 2018, 18:08:52 UTC |
1ee4316 | alex@thinkpad | 30 January 2018, 18:07:57 UTC | Makefile: prevent install_check from running in parallel with building regular ML (conflicts while compiling from the same sources - http://www.magiclantern.fm/forum/index.php?topic=991.msg194315#msg194315 ) | 30 January 2018, 18:07:57 UTC |
d4657cc | alex@thinkpad | 29 January 2018, 21:01:52 UTC | 70D: mpu_send/recv stubs | 29 January 2018, 21:01:52 UTC |
a6d890e | alex@thinkpad | 29 January 2018, 20:57:56 UTC | Merged 100D_merge_fw101 into dm-spy-experiments | 29 January 2018, 20:57:56 UTC |
158a93b | alex@thinkpad | 29 January 2018, 20:45:36 UTC | Merged 70D_merge_fw112 into dm-spy-experiments | 29 January 2018, 20:45:36 UTC |
57734c1 | alex@thinkpad | 29 January 2018, 20:41:26 UTC | Merged patchmgr into dm-spy-experiments | 29 January 2018, 20:41:26 UTC |
b923894 | alex@thinkpad | 29 January 2018, 20:40:10 UTC | Patch manager: nicer way to avoid calling _reapply_cache_patches when not present (fixes compiling the installer target and removes the need for defining NO_CACHE_PATCHES) | 29 January 2018, 20:40:10 UTC |