Any C++ developers here?

We may earn a small commission from affiliate links and paid advertisements. Terms

lsvtec

GNU/Linux Evangelist
I have an STL question, what would cause the vector.begin( ) method to SIGSEGV? (I am sure that the vector in question was constructed properly)
 
if you are asking this question you got me beat. i dont know how vectors work, but i do know they automanage memory, so getting a segmentation fault makes no sense.
 
That is mostly true, you can still hose them by overrunning a buffer into their memory accounting areas, but I have checked that already...

To make the problem even more strange it seems to be limited to certain hardware. Running SLES, RHEL, Gentoo, Debian, or Ubuntu on x86 works fine, running RHEL or SLES on an IBM POWER 720 works fine, but it breaks on any distro running on an IBM POWER Blade JS20.
 
hmmm....

you might have answered your own question here. I dont know if the Power 720 is an x86_64 based system but i KNOW the JS20 is. This might be yet another issue with the 64bit architecture.
 
Actaully they are both runnung POWER processors. The 720 runs 4 POWER 5 dual cores, and I am not sure if the JS20 is POWER 5 or POWER 4.

The app is compiled as a 32 bit app (POWER does the 32 bit and 64 bit concurrently better than AMD because it doesn't have all that legacy x86 cruft).
 
thats just plain confusing the JS20 uses dual POWER 5s too. but the app runs as planned on all hardware except the js20?

have you tried it on a different js20 blade. might be running into bad memory (although restarting will usually solve this issue).
 
Yes, the same problem occurs on the two different JS20 Blades I have used.I have install SLES 9 gm and SP3, SLES 10 gm, and RHEL 4 ga. All of these exhibit the same SEGV.
 
god thats wacky. the only thing i can think of is hardware. maybe a wacky south bridge or something...

how much memory do the js20s have? you might just be overloading the ram.
 
Back
Top