|
From: Christian E. (gmx) <Ch....@gm...> - 2003-09-03 13:32:06
|
Hi I've a C - programm which accesses the sg-driver. This program works ok when I test it with valgrind. When I run it without valgrind, I've the problem that there is a mysterious errror (SCSI-timeout) when I want to read ~1MB from a scsi-device (sometimes later sometimes earlier). So I wanted to know if maybe valgrind did something which avoids this error... maybe some aligning or so. Thx, Christian |
|
From: Joerg B. <jo...@we...> - 2003-09-03 13:49:43
|
Christian Ehrlicher (gmx) wrote:
> Hi
>
> I've a C - programm which accesses the sg-driver. This program works ok
> when I test it with valgrind. When I run it without valgrind, I've the
> problem that there is a mysterious errror (SCSI-timeout) when I want to
> read ~1MB from a scsi-device (sometimes later sometimes earlier). So I
> wanted to know if maybe valgrind did something which avoids this
> error... maybe some aligning or so.
valgrind for sure changes the timing of a running program - usualy
it slows the program a bit down. Could this have the effect you see?
Joerg
|
|
From: Paul H. <pa...@ha...> - 2003-09-03 14:54:20
|
On Wed, 3 Sep 2003, Christian Ehrlicher (gmx) wrote: > Hi > I've a C - programm which accesses the sg-driver. This program works ok > when I test it with valgrind. When I run it without valgrind, I've the > problem that there is a mysterious errror (SCSI-timeout) when I want to > read ~1MB from a scsi-device (sometimes later sometimes earlier). So I > wanted to know if maybe valgrind did something which avoids this > error... maybe some aligning or so. Trace the system calls with strace, both with and without valgrind. Look at the system calls that access the problem device. > Thx, > Christian > > > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > Valgrind-users mailing list > Val...@li... > https://lists.sourceforge.net/lists/listinfo/valgrind-users > |
|
From: Nicholas N. <nj...@ca...> - 2003-09-03 15:13:43
|
On Wed, 3 Sep 2003, Paul Haas wrote: > Trace the system calls with strace, both with and without valgrind. Look > at the system calls that access the problem device. You can use --trace-syscalls=yes to get Valgrind to trace the syscalls itself, too. N |
|
From: Christian E. (gmx) <Ch....@gm...> - 2003-09-04 09:04:40
|
Hello The idea with the timings was right - I insert a sleep(1) before each read and it worked. But I wonder why this doesn't happen with the other settings... I'll try to trace the syscalls when I've got time. Thx for your help :-) Christian Ehrlicher |