On Sat, Jul 22, 2006, Paul J Stevens <[EMAIL PROTECTED]> said: > Aaron Stone wrote: >> On Fri, Jul 21, 2006, Paul J Stevens <[EMAIL PROTECTED]> said: >> >>> Aaron Stone wrote: >>>> On Fri, 2006-07-21 at 16:59 +0200, Paul J Stevens wrote: >>>>> I'm sorry Aaron, but I'm rejecting your patch. It starts breaking stuff >>>>> (sig11) all over the place, I don't have time to fix them all, and I >>>>> don't want to leave the tree in a non-functional state. I'm leaving the >>>>> country next week, and I would like to get 2.1.7 out before then. >>>> Bah, crap. I'll see about un-segfaulting in the next 24 hours. >> >> Paul, I don't know what segfaults you're seeing, but I'm not seeing them. >> Please point me in the direction of the problems you're getting so that I >> can fix them! > > static build, --with-mysql --with-auth-ldap --with-sieve > using libsieve-2.1.11 > > ==9317== Process terminating with default action of signal 11 (SIGSEGV) > ==9317== Access not within mapped region at address 0xDEA44 > ==9317== at 0x419C0F9: vasnprintf (in /usr/lib/libsieve.so.1.0.5) > ==9317== by 0x419B662: vasprintf (in /usr/lib/libsieve.so.1.0.5) > ==9317== by 0x42E579A: g_vasprintf (in /usr/lib/libglib-2.0.so.0.1000.2) > ==9317== by 0x42D8673: (within /usr/lib/libglib-2.0.so.0.1000.2) > ==9317== by 0x42D86C1: g_string_printf (in > /usr/lib/libglib-2.0.so.0.1000.2) > ==9317== by 0x8050A95: _header_cache (dbmail-message.c:947) > ==9317== by 0x42DC63B: g_tree_foreach (in > /usr/lib/libglib-2.0.so.0.1000.2) > ==9317== by 0x805072B: dbmail_message_headers_cache > (dbmail-message.c:857) > ==9317== by 0x805043F: dbmail_message_store (dbmail-message.c:778) > ==9317== by 0x8063427: insert_messages (pipe.c:581) > ==9317== by 0x804C6A9: test_insert_messages (check_dbmail_deliver.c:109) > ==9317== by 0x806DB81: srunner_run_all (in > /home/paul/git/dbmail-2.1/check_dbmail_deliver) > ==9317== > > It took me a while to realize that the vasnprintf call are being handled > by libsieve. Now why are you providing gnu libc functions in sieve?
It's from gnulib, so that I get a consistent (read: doesn't crash on null pointers) vasnprintf for debugging purposes. I'm surprised that the libraries linked like this; the glibc symbol for vasnprintf must be weaker than the libSieve symbol. I'll push out a fixed libSieve tomorrow that moves this into the libsieve_ prefixed namespace. Aaron
