On Wed, Jul 22, 2015 at 10:49:38AM -0300, Pedro Arthur wrote:
> Last week I worked on fixing the following
> * Memory allocation checks
> * Code style
> I also investigated the ffplay bug (./ffplay -f lavfi testsrc -vf
> format=yuvj420p) which presents a grey screen. It still not fixed in this
> patch but I already found the cause and I intend to fix in the next patch.
> 
> For the next week I intend to write some documentation for the code added
> and start to work on the vertical scaler.

i suggest to try to fix all bugs before adding more features. Adding
docs and any other work should be no problem but its better if
code is in fully testable condition early so we could for example
use git bisect and not be stuck with a range of untestable revissions

Here are 2 more cases that dont work:
./ffplay -f lavfi testsrc  -vf format=yvyu422,scale=flags=1     (black screen)
valgrind shows this:

==4194== Invalid read of size 1
==4194==    at 0xE4499B: ff_hyscale_fast_mmxext 
(hscale_fast_bilinear_simd.c:278)
==4194==    by 0xE7D82C: lum_h_scale (slice.c:164)
==4194==    by 0xE38C3A: swscale (swscale.c:530)
==4194==    by 0xE3B54A: sws_scale (swscale.c:1177)
==4194==    by 0x4BF5FE: scale_slice (vf_scale.c:440)
==4194==    by 0x4BFC69: filter_frame (vf_scale.c:540)
==4194==    by 0x446E6E: ff_filter_frame_framed (avfilter.c:1091)
==4194==    by 0x4473E6: ff_filter_frame (avfilter.c:1172)
==4194==    by 0x4469D1: default_filter_frame (avfilter.c:1002)
==4194==    by 0x446E6E: ff_filter_frame_framed (avfilter.c:1091)
==4194==    by 0x4473E6: ff_filter_frame (avfilter.c:1172)
==4194==    by 0x4BFC8B: filter_frame (vf_scale.c:544)
==4194==  Address 0x20c643ff is 1 bytes before a block of size 720 alloc'd
==4194==    at 0x4C2A6C5: memalign (vg_replace_malloc.c:727)
==4194==    by 0x4C2A760: posix_memalign (vg_replace_malloc.c:876)
==4194==    by 0xEB84BB: av_malloc (mem.c:97)
==4194==    by 0xE7CE85: alloc_lines (slice.c:27)
==4194==    by 0xE7E9BB: ff_init_filters (slice.c:465)
==4194==    by 0xE428A0: sws_init_context (utils.c:1603)
==4194==    by 0x4BF266: config_props (vf_scale.c:392)
==4194==    by 0x444A3E: avfilter_config_links (avfilter.c:262)
==4194==    by 0x4449D2: avfilter_config_links (avfilter.c:251)
==4194==    by 0x447B6B: graph_config_links (avfiltergraph.c:275)
==4194==    by 0x44AA20: avfilter_graph_config (avfiltergraph.c:1212)
==4194==    by 0x41EAF3: configure_filtergraph (ffplay.c:1951)
...


./ffplay -f lavfi testsrc  -vf format=rgb555 (gray and some odd distortion)
the issue goes away with -cpuflags 0, so it seems related to the use
of asm code
valgrind shows this:
==4069== Thread 17:
==4069== Invalid read of size 8
==4069==    at 0xE4B47B: yuv2rgb555_1_mmxext (swscale_template.c:1328)
==4069==    by 0x1A37B5EF: ???
==4069==    by 0x1A37B84F: ???
==4069==    by 0xE39408: swscale (swscale.c:699)
==4069==    by 0x13F: ???
==4069==  Address 0x20c50c80 is 16 bytes after a block of size 720 alloc'd
==4069==    at 0x4C2A6C5: memalign (vg_replace_malloc.c:727)
==4069==    by 0x4C2A760: posix_memalign (vg_replace_malloc.c:876)
==4069==    by 0xEB84BB: av_malloc (mem.c:97)
==4069==    by 0xE7CE85: alloc_lines (slice.c:27)
==4069==    by 0xE7EAA9: ff_init_filters (slice.c:470)
==4069==    by 0xE428A0: sws_init_context (utils.c:1603)
==4069==    by 0x4BF266: config_props (vf_scale.c:392)
==4069==    by 0x444A3E: avfilter_config_links (avfilter.c:262)
==4069==    by 0x4449D2: avfilter_config_links (avfilter.c:251)
==4069==    by 0x4449D2: avfilter_config_links (avfilter.c:251)
==4069==    by 0x4449D2: avfilter_config_links (avfilter.c:251)
==4069==    by 0x447B6B: graph_config_links (avfiltergraph.c:275)


also if you need help with debuging something / are stuck with
debuging, dont wait the whole week but please mail me or the list
or try IRC
iam happy to help!

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

The real ebay dictionary, page 2
"100% positive feedback" - "All either got their money back or didnt complain"
"Best seller ever, very honest" - "Seller refunded buyer after failed scam"

Attachment: signature.asc
Description: Digital signature

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Reply via email to