[prev] [thread] [next] [lurker] [Date index for 2006/12/11]
On Dec 11, 2006, at 1:58 AM, Yossi Kreinin wrote: > Right. That's why it fails when I pass the module. What I would like > to know is (1) why can't gcc on RHEL work with an ELF shared object > compiled on SuSE, although nm, objdump & dlopen/dlsym on RHEL all > agree that the object is kosher and (2) which function got which > invalid argument. Telling that some argument to some function is wrong > is less than helpful. I understand. Proper error handling is hard, and it's hateful to have to fire up truss or equivalent to see what fucked up and track it down, but this is only a level 3 hate. If Microsoft were doing it the same failure would produce a dialog box telling you that the disk is full, and see your system administrator. Your random passive-aggressive UNIX vendor would have back-ported their mainframe compiler to UNIX and it would either spit out a misleading stack trace from an unrelated part of their runtime, or take the "ed" philosophy to heart and simply produce "badmodule.o?". If Knuth was doing it that'd be expanded to ")))))))))))))(((((()))))(badmodule.o)[...]badmodule.o?". > Right. They made me implement a Unix-like shell once, so I happen to > know the sacred protocols of Unix job control, so I've guessed the > scenario you suggest. I couldn't care less as a user though. As a > user, I like the behavior of tcsh@SuSE (and all other shells I've > used) - saying that it killed something right after it did so. Why is > tcsh@RHEL different? Probably because they compiled it with the options that said "I want a thin shell that doesn't do stuff like busywaiting after certain commands to see if there's a delayed result" instead of "I want a thick shell that isolates a few of the details of UNIX from me". Depending on how thick the shell is, both options are hateful for different people at different times. > Right. I know that. I know that it can look whatever it likes because > that's the way they distribute it and still have Linux somewhere in > it's guts. I'm not saying that either RHEL or SuSE should avoid using > the name "Linux". I'm saying that since they ship two similar things, > it would be very /nice/ of them to avoid introducing inconsistencies > for the sake of it. And the piece of shite called tcsh should lay > where it was dropped the first time someone decided to litter a > storage device with it's copy. That would be "$PREFIX/bin", I hope. > I don't care where tcsh is. The scripts running it with #!/bin/tcsh or > the like do care, Anything that causes pain when using scripts running it with #!/*bin/tcsh is good. Writing scripts in tcsh is so hateful that anything that hastens the day when hordes of angry users descend on the programmers who did it and string them up by their own guts should be heartily encouraged.There's stuff above here
Generated at 22:02 on 27 Dec 2006 by mariachi 0.52