On Monday 10 October 2005 05:19 am, Joerg Schilling wrote: > So debugging on Linux is what Brian Cantrill calls sending a new debug > airplane with 300 new passengers over the ocean in hope that it will crash > too just in oder to find the reason instead of checking the rests of the > one machine that did just crash.
Not a bad analogy, IMO. It's fairly ironic that folks are always quick to defend Linux for some of the shortcomings that developers run into. The very lack of a lightweight process is yet another example. I hear that's possibly fixed now, but for me Linux is like "The Boy Who Cried Wolf", I just couldn't stick with it and continue to run in such an ad hoc environment, so even if it doesn't clone processes today, I had given up on it long before it did. For the most part, all Linux systems are debug systems. At Sun we run a server with the latest builds and yes, we have problems, and the engineers fix them. There is no better environment than having 800 users hammer on the server, NFS mounting their home directories and relying on it for them mail. This shakes out quite a bit of bugs, and makes the product much better. > I believe it is unacceptable to be forced to compile a special debug binary > just in order to be able to use a debugger. Specially when it is known that > GCC had bugs when compiling with "-O2" compared to "-O2 -g". I don't think it's unacceptable to be forced to compile a special debug binary, most systems have been like that for ages. What is unacceptable is that you pretty much *HAVE* to compile a special debug binary to figure out what is going on...this reminds me of a time I was working on Embedded Linux, which is does ok at, IMO, but I couldn't get shared memory working. So, the response I got was, "well you have the source". Ok, so I go in and figure out that the kernel was broken, and spend 2 full days doing it. Sure, it's good to have the source in those cases, but it's not as if you're so blessed to have to go into it. Things like shared memory were debugged years ago on Solaris. -- Alan DuBoff - Sun Microsystems Solaris x86 Engineering _______________________________________________ opensolaris-discuss mailing list [email protected]
