valgrind sucks

[Asher]

The official compiler on AIX these days is xlc for C, xlC for C++.

gcc is supported but only if you go get the ****er yourself.

[Asher]

Note xlc/xlC costs $3600 USD to buy. You pay a pretty penny for it to tell you your code is stupid.

[BlackCat]

Sorry for not remembering the right name - it's some 10 years since I had the "pleasure" to battle that stupid system

[Kuciwalker]

Bad code. Both the C/C++ Standards demand main() have a return type.

[Asher]

Bad code. Both the C/C++ Standards demand main() have a return type.

I am very serious.

[Asher]

Sorry for not remembering the right name - it's some 10 years since I had the "pleasure" to battle that stupid system

Fact: xlc/xlC are the fastest C/C++ compilers in the world. They smoke gcc so bad it's not even funny.

[BlackCat]

Note xlc/xlC costs $3600 USD to buy. You pay a pretty penny for it to tell you your code is stupid.

Well, they may have improved, but I had a program that I could compile and run on a nuber of unix flavors except AIX. It compiled fine without warns but crashed before main.

[Asher]

http://www.macresearch.org/science_on_a_g5_ps3_macpro_benchmarks

[Asher]

Well, they may have improved, but I had a program that I could compile and run on a nuber of unix flavors except AIX. It compiled fine without warns but crashed before main.

Post 2001 or so they became very anal about standards compliance.

[BlackCat]

Could explain it - we scrapped our last AIX before 2000.

Though, it still pisses me off that it crash before main - makes it kinda problematic to debug .

[Kuciwalker]

main ()
{
	struct sembuf sbuf;
	int semid;

	semid = semget(KEY, 1, IPC_CREAT | IPC_EXCL | 0777);
	printf("semid = %d\n", semid);

	semctl(semid, 0, SETVAL, 0);

	semctl(semid, 0, IPC_RMID);
	puts("destroyed semid");
}

This results in no valgrind errors. But the following does:

main ()
{
	struct sembuf sbuf;
	int semid;
	int r;

	semid = semget(KEY, 1, IPC_CREAT | IPC_EXCL | 0777);
	printf("semid = %d\n", semid);

	r = semctl(semid, 0, SETVAL, 0);
	printf("initialized semid = %d\n", r);

	semctl(semid, 0, IPC_RMID);
	puts("destroyed semid");
}

==13830== Syscall param semctl(arg) points to uninitialised byte(s)
==13830==    at 0xB8AA08: semctl@@GLIBC_2.2 (in /lib/tls/libc-2.3.4.so)
==13830==    by 0x804849E: main (in /home/a0132838/play/semtest)
==13830==  Address 0xBEE123D4 is on thread 1's stack

What the fvck.

[Asher]

Never used this lib.

[BlackCat]

Valgrind likes to tease Kuci

[Kuciwalker]

xpost

Be glad. I'm tempted to take the codebase I'm working with and just migrate it wholesale to named semaphores in semaphore.h. sys/sem.h dates back to 1983.

[Jon Miller]

You have never found a compiler bug before?

JM