I could use some pointers on how to debug a segfault in ld.so on
i386 -current (as of last night). I'm trying (so far in vain) to figure out
a
problem with devel/jdk/1.2 in -current. javac segfaults before it reaches
the main function. niklas@ gave me some pointers on how to track down
problems in ld.so, but I've kind of hit the wall and can't make any more
progress...

Using DL_DEB statements in the _dl_boot function in libexec/ld.so :
loader.c, I know that the segfault is happening while executing line
365/366:

(*(void (*)(Elf_Addr))(sym->st_value + ooff))
((Elf_Addr)_dl_dtors);

The comments and code indicate this code calls the atexit function. I placed
printf statements in the atexit function of libc/stdlib/atexit.c (don't do
this at home), however I never reach them while attempting to execute
javac_g. (I did see my printf all over while executing other stuff and I
would not recommend doing it at least not without a backup of libc).

OK, at this point I'm a bit perplexed because I expected to see the printf's
I added to libc/stdlib/atexit.c, but I didn't. However, I know something is
being executed by lines 365/366 in loader.c because DL_DEB statements are
being output from dlopen:

calling atexit - (before line 365 of loader.c)
dlopen: loading: libc.so
linking dep /usr/lib/libc.so.30.0 as child of
/usr/ports/devel/jdk/1.2/w-jdk-1.2.2/build/bsd/bin/i386/green_threads/javac_
g
dlopen returning: already loaded (at line 72 of dlfcn.c)

Program received signal SIGSEGV, Segmentation fault.
0x0 in ?? ()
(gdb)

So I'm stuck. Something is being executed but I don't know what. I haven't
been able to put a breakpoint in using gdb because all of this happens
before the main function of javac_g. Is there a way I can halt execution and
step through this code? or Would someone enlighten me as to what is being
executed by line 365/366 so I can continue using DL_DEB or printf's to
figure this out? Any assistance, RTFM or flames welcome ;-)

Thanks,
-Kurt