|
From: Nicholas N. <nj...@cs...> - 2008-04-20 22:48:32
|
On Sun, 20 Apr 2008, Bart Van Assche wrote: > On Thursday April 17 Eugene Toder filed a Valgrind bug report because > there was not yet a mallinfo() implementation in Valgrind. The first > patch attached to the bug report looked to me like an easy to maintain > basic mallinfo() implementation. I verified the implementation against > the behavior documented in the glibc manual and the Solaris mallinfo() > man page, and I also verified with Google Code Search that the > implemented behavior is what is expected by some mallinfo() users. > Since the implementation looked OK to me, I added it on the trunk. > > By this time Eugene was kind enough to supply a full mallinfo() > implementation and regression test. Since there is no recent standard > defining the mallinfo() behavior, and since it's not clear to me how > mallinfo() behaves on e.g. AIX, I decided to throw this issue on the > Valgrind developers mailing list. What is the opinion of the other > Valgrind developers about adding a full mallinfo() implementation to > Valgrind ? It seems ok to me. People have asked for this before. IIRC he only bothered to fill in the field that holds the total number of bytes, no? Since it's rare and poorly specified I'm not that worried by what Valgrind does... N |