Re: Assertion in sort.c

[Ben Pfaff]

John Darrington <address@hidden> writes:

> Although big-input-2.sh now passes, running it under valgrind shows
> that there may be some problems lurking still:
>
>
> ==19431== Syscall param write(buf) contains uninitialised or unaddressable 
> byte(s)
> ==19431==    at 0x1BC06B38: write (in /usr/lib/debug/libc-2.3.2.so)
> ==19431==    by 0x807602B: flush_buffer (casefile.c:334)
> ==19431==    by 0x80761C3: casefile_to_disk (casefile.c:373)
> ==19431==    by 0x8075E67: casefile_append (casefile.c:296)
> ==19431==  Address 0x1C4A8652 is 2 bytes inside a block of size 8192 alloc'd
> ==19431==    at 0x1B904EDD: malloc (vg_replace_malloc.c:131)
> ==19431==    by 0x8070F66: xmalloc (alloc.c:39)
> ==19431==    by 0x807610B: casefile_to_disk (casefile.c:364)
> ==19431==    by 0x8075E67: casefile_append (casefile.c:296)

That's okay.  The casefile code doesn't always initialize the
entire buffer it writes to disk, but in that case it ignores
those bytes when it reads them back from disk.  It's generally
due to padding bytes between string variables added for alignment
reasons, and you can happily ignore it.
-- 
"In this world that Hugh Heffner had made,
 he alone seemed forever bunnyless."
--John D. MacDonald