Changes between Version 23 and Version 24 of zzuf/bugs

Timestamp:

01/31/10 14:59:53 (3 years ago)

Author:

sam

Comment:

testsuite details

zzuf/bugs

@@ -29 +29 @@
 || '''WMV''' || '''SIGSEGV''' || '''SIGSEGV''' || '''SIGSEGV''' || ? || '''SIGSEGV''' ||
 || '''FLV''' || '''SIGSEGV''' || '''SIGSEGV''' || '''SIGSEGV''' || ? || '''SIGSEGV''' ||
+
+=== Testing protocol ===
+
+The zzuf commands used to find bugs were rather straightforward. A few flags are recurrent:
+ * `-r0.0001:0.02` to try several fuzzing ratios
+ * `-s0:10000` to stop after 10000 tries
+ * `-b8` to skip the first 8 bytes and ensure that the file format is not misinterpreted
+
+MPlayer is very easy to test, thanks to its `-benchmark` flag:
+{{{
+#!sh
+% zzuf -vq -c -r0.0001:0.02 -s0:10000 -b8- -j8 -T5 -S mplayer -benchmark \
+           -ao pcm:file=/dev/null -vo md5sum:outfile=/dev/null filename
+}}}
+
+VLC doesn’t provide a benchmark flag, forcing us to spawn a lot more parallel processes using `-j`, so that the test goes a bit faster:
+{{{
+#!sh
+% zzuf -vq -c -r0.0001:0.02 -s0:10000 -b8- -j80 vlc -I dummy \
+           -A dummy -V dummy filename vlc://quit
+}}}
+
+GStreamer is as easy to test as MPlayer:
+{{{
+#!sh
+% zzuf -vq -I'orig.*' -r0.0001:0.02 -s0:10000 -b8- -j5 -S gst-launch-0.10 \
+           filesrc location=filename '!' decodebin '!' fakesink
+}}}
+
+I could not find a benchmark mode for xine, so I tested it using the libcaca output:
+{{{
+#!sh
+% CACA_DRIVER=raw zzuf -vq -c -r0.0001:0.02 -s0:10000 -b8- -j15 -S cacaxine -A none -q filename
+}}}
+
+Finally, mpg321 and ogg123 don’t have a benchmark mode either. We increase the `-j` value as well:
+{{{
+#!sh
+% zzuf -vq -c -r0.0001:0.02 -s0:10000 -b8- -j10 mpg321 filename
+% zzuf -vq -c -r0.0001:0.02 -s0:10000 -b8- -j10 ogg123 -d null filename
+}}}
 
 == The January 2007 media player debacle ==