Description:
This symptom seems to be new in 5.0.72sp1,
I see corresponding packages for the previous versions.
For all configurations, the compiler crashes on module "mysys/mf_format.c".
I have been able to isolate it to this part:
> diff -C 6 mf_format.c.ORI mf_format.c
*** mf_format.c.ORI Tue Jan 13 15:42:14 2009
--- mf_format.c Thu Jan 15 22:44:33 2009
***************
*** 94,113 ****
--- 94,123 ----
(void) strmov(pos,ext); /* Don't convert extension */
}
/*
If MY_RETURN_REAL_PATH and MY_RESOLVE_SYMLINK is given, only do
realpath if the file is a symbolic link
*/
+
+ /* ERROR on HP-UX 11.11 using the native compiler version aCC: HP ANSI C++ B3910B A.03.65
+ in 64 bit mode with flag +O2:
+ ANY condition below, even the trivial ( (1 + 2) == 3 )
+ controlling a non-empty "then" or "else" branch,
+ will cause the compiler to crash internally with signal 11.
+ */
if (flag & MY_RETURN_REAL_PATH)
+ {
(void) my_realpath(to, to, MYF(flag & MY_RESOLVE_SYMLINKS ?
MY_RESOLVE_LINK: 0));
+ }
else if (flag & MY_RESOLVE_SYMLINKS)
{
strmov(buff,to);
(void) my_readlink(to, buff, MYF(0));
}
+
DBUG_RETURN(to);
} /* fn_format */
/*
strlength(const string str)
This is the compiler call:
source='mf_format.c' object='mf_format.o' libtool=no \
DEPDIR=.deps depmode=hp /bin/sh ../depcomp \
cc -DDEFAULT_BASEDIR=\"/usr/local/mysql\" -DDATADIR="\"/usr/local/mysql/data\"" -DDEFAULT_CHARSET_HOME="\"/usr/local/mysql\"" -DSHAREDIR="\"/usr/local/mysql/share/mysql\"" -DDEFAULT_HOME_ENV=MYSQL_HOME -DDEFAULT_GROUP_SUFFIX_ENV=MYSQL_GROUP_SUFFIX -DDEFAULT_SYSCONFDIR="\"/usr/local/mysql/etc\"" -DHAVE_CONFIG_H -I. -I../include -I../zlib -I../include -I../include -I. -DDBUG_OFF -g +O2 +DD64 -DHPUX11 -DSNPRINTF_RETURN_TRUNC -DHAVE_BROKEN_PREAD -DDONT_USE_FINITE -DHAVE_BROKEN_GETPASS -DNO_FCNTL_NONBLOCK -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT -DHAVE_BROKEN_INLINE -c mf_format.c
cc: error 1405: "/tmp/ansic/lbin/ccom" terminated abnormally with signal 11.
*** Error exit code 9
How to repeat:
Try to build it.
Suggested fix:
Check the differences between 5.0.72 (worked), 5.0.72sp1 (fails), and 5.0.74 (worked).
When I created the sources for 5.0.72sp1, I took 4 showstopper bugs from 5.0.74 and backported them onto the 5.0.72 sources, all these merges succeeded without any issues.
Also, I am not aware of any significant build or test issues on any other platform (yet).