On 1/16/06, James Vega <[EMAIL PROTECTED]> wrote: > I uploaded 1.19.0 to Debian this weekend (yeah, I've been lazy) and > noticed some a build failure[0] on ia64.
Hey, it's been christmas. That's as good an excuse as any. > > Running fish_tests during "make test" segfaulted with the following > backtrace: > > (gdb) bt > #0 0x4000000000069040 in hash_wcs_func (data=0x140f0) at util.c:506 > #1 0x4000000000066a80 in hash_search (h=0x60000000000149a0, key=0x140f0) at > util.c:222 > #2 0x4000000000067470 in hash_put (h=0x60000000000149a0, key=0x140f0, > data=0x60000000000140d0) > at util.c:294 > #3 0x4000000000029e20 in env_set (key=0x60000000000140a0, > val=0x60000000000140b4, var_mode=6) > at env.c:649 > #4 0x4000000000028280 in env_init () at env.c:336 > #5 0x4000000000093720 in main (argc=1, argv=0x60000fffff95bc68) at > fish_tests.c:680 > > Notice that the key being passed to env_set (0x60000000000140a0) is > truncated when it's used in the hash_put call (0x140f0). The call to hash_put should use a duplicated version of the string sent to env_set. As you say, the memory adress seems to be truncated. I'm guessing that sinse wcsdup is undeclared, it is assumed to return and int, when it really returns a pointer. I'm stupid. But hey, it was christmas! > > The same backtrace occurs when starting up fish, as expected, since this > is during env_init. No surprise. I'd do a 1.19.1 fixing this, but I really think 1.20.0 is ready, and it contains fixes for several other bugs, as well as i18n support, the new syntax for the and/or builtins, completions for gpg, correct completions for a few commands on BSD and lots of other features, so I think I'll try to release that today instead. > > I'd also like to point out that there are a number of files[1] with "implicit > declaration of function ..." warnings. I'll be glad to help check for > stuff like this before releases, but I'd also recommend a stricter > pre-release build process. Maybe adding a 'debug' rule to the Makefile > that adds the following CFLAGS: -O0 -Wno-unused -Werror -g Very sane idea. I think you've suggested this before, but it sliped my mind. Now I've done exactly what you suggested, hopefully this will help. > > I just tested the latest Darcs tree and these problems seem to be fixed, > but I think it should be a goal to not have out-of-the-box build > failures of a fresh release and I'm willing to do what I can to help out > with that. The recent Darcs tree also seems to have the fewest pointer > <-> int conversion warnings I've seen. :) Looks like things are > working themselves out. Mostly, I think _you_ are working them out. > > On another note, there's a slight typo in INSTALL which I've attached a > patch for. I corrected the same mistake before going to bed last night. But thanks for the patch! > > James > [snipped footnotes] -- Axel ------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK! http://ads.osdn.com/?ad_idv37&alloc_id865&op=click _______________________________________________ Fish-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/fish-users
