Unless *both* the samples *and* the md5 checksum were in bad memory.
Of course, --verify would catch this, which just says that anyone who
ever gets an MD5 error on their system should always use --verify in
the future.
On Jul 27, 2007, at 10:50, Josh Coalson wrote:
--- Harry Sack <[EMAIL PROTECTED]> wrote:
2007/7/27, Josh Coalson <[EMAIL PROTECTED]>:
But how is it possible then the FLAC encoder allows files which have
a bad
resulting MD5 to be encoded? Is it because of the bad ram, ... this
incorrect MD5 is not detected during encoding?
it happens like this, x.wav gets encoded with flac.exe on a machine
with bad ram. flac.exe reads some audio data from x.wav into memory
and encodes it. samples are buffered to md5 checksummer which writes
intermediate checksum to bad memory. md5 is corrupted but samples
were encoded ok. the user will not know unless the --verify option
was used during encoding.
Is it also possible the samples themself get corrupted, because they
are too (just like the md5) written to the memory for encoding them?
yes, but in that case when you decode it you won't get an md5
error. from flac's point of view everything went fine because
it can't know the samples coming in were wrong to begin with.
_______________________________________________
Flac mailing list
Flac@xiph.org
http://lists.xiph.org/mailman/listinfo/flac
