On Sun, Aug 21, 2005 at 01:45:37PM -0400, Justin Pryzby wrote: > > hmm - aside from the cosmetic issue you reported with libXt (an overlapping > > memcpy is just that), what other warnings are you seeing with valgrind? > Right; its cosmetic because the obvious implementation of memcpy is: > > for ( ; n; n--) { > *dest++ = *src++; > } > > However, the manpage _says_: "The memory areas should not overlap.", > and so it is probably best to make no assumptions about its > implementation. true. But in this case, valgrind is just pointing out that the code is inefficient. (I don't recall what I was working on when I fixed it in libXt, but it was something that valgrind reported then). > Invalid reads and writes. I've attached the output of `valgrind > --log-file=xterm-valgrind ./xterm`, resizing the xterm to LINES=1, and > running top -d.3. > > > (and of course, what version of xterm, > We should probably have a bugscript to include xterm's version in > reports.. cloning now. "xterm -v" gives the information (and it doesn't try to open the display ;-) > It appears to be: > > ++<LI><A HREF="#xterm_200">Patch #200 - 2005/2/6 - XFree86 4.4.99.23</A> > > > what size is "super-small", etc). > LINES=1. In fact, I just ran `echo $LINES` in a 1 line xterm, and it > crashed (or, something did, anyway). (It didn't crash under valgrind, > though). yes - that's expected behavior. A 1-line xterm "can't" scroll, since scrolling regions have to be at least 2 lines. The earlier bug fix exposed this hole. I do use valgrind - the last I recall is that the only unaccounted problems are in glibc when the locale isn't POSIX. -- Thomas E. Dickey http://invisible-island.net ftp://invisible-island.net
Attachment:
pgpv6tERc3hXu.pgp
Description: PGP signature