[argyllcms] Re: valgrind memcheck on profile

Guy K. Kloss wrote:
To get a better overview what profile is doing, I've created a calltree using "valgrind --tool=callgrind". Doing that (just for the fun of it) I also ran profile with the memcheck tool. There were (only very few) "weak spots", but it may help to identify and resolve them. Especially (in my case here, profiling an RGB camera input device) one memory boundary violation has been encountered (by 11 bytes). 

It's not clear to me from the valgrind output, but I suspect
that in fact this is one byte that is uninitialized (byte 11 of the
Lut Tag), rather than being a memory boundary violation by 11 bytes.
Someone actually pointed this out to me a week or two ago (funny
how stuff like this lies around for years, and is then discovered
at almost the same time).

> ==30053==    by 0x8064582: icmLut_write (icc.c:5941)

> ==30053==  Address 0x459bf53 is 11 bytes inside a block of size 41,818 alloc'd

My current code sets bp+11 to zero in icc.c/icmLut_write()


==30053==    by 0x804CC03: make_input_icc (profin.c:362)


Yep, tpat needs freeing at the end of profin.c/make_input_icc
and profout.c/make_output_icc

Thanks for the report.

Graeme Gill.

Other related posts: