|
Sean Boudreau(deleted)
11/25/2008 1:22 PM
post17295
|
On Tue, Nov 25, 2008 at 01:07:53PM -0500, Mario Charest wrote:
> _cmdname() doesn't behave the same under 6.3.2 and 6.4 (the same binary build under 6.3.2 )
>
> When run from a machine called value
>
> under QNX 6.3.2 :
>
> cd /net/neutrino/u/mcharest/logc/bin
> ./testApp
> _cmdname() returned ./testApp.
>
> under QNX6 6.4.0
>
> cd /net/neutrino/u/mcharest/logc/bin
> ./testApp
> _cmdname() returned /net/value/net/neutrino/u/mcharest/logc/bin/.testApp
>
>
> This is causing grief to our code any way to get the same behavior. I'm surprise it's different since it is using
libc.so.2 and not libc.so.3
>
Proc sets this up when loading the process so it's not
a libc issue. This looks to have been done on purpose
and it now matches how it's doc'd as behaving.
Regards,
-seanb
|
|
|
|
|
|
Brian Stecher
11/25/2008 2:45 PM
post17304
|
On Tue, Nov 25, 2008 at 01:07:48PM -0500, Mario Charest wrote:
> _cmdname() doesn't behave the same under 6.3.2 and 6.4 (the same binary build under 6.3.2 )
As Sean said, this was a bug fix - _cmdname() needs to always return the
full path so that it's a valid file name to open even if the application does
a chdir().
> This is causing grief to our code any way to get the same behavior. I'm surprise it's different since it is using
libc.so.2 and not libc.so.3
There's no way to get the old behaviour. What kind of grief is it causing
you?
BTW, libc.so.2 is pretty much a shell with just GCC compiler support
routines defined within it. Even if you run a 6.3.2 binary that wants to
use libc.so.2, the vast majority of the functions being invoked are actually
in libc.so.3.
|
|
|
|
|
