pthreads, fork & exec vs. system(), solaris, others (fwd)

[Kevin D. Clark]

"Steven W. Orr" <steveo at syslang.net> writes:

> A friend wrote me with a problem. The basic gist is that it works
> under Solaris but fails under AIX. Description below. Any takers?

Generally speaking, the Solaris environment is very friendly.  The AIX
environment isn't very friendly.

I'm not even making a pejorative statement here -- sometimes incorrect
code runs "just fine" on Solaris.  On AIX the same code will probably
blow up.  Myself, I like to know when such problems exist, so
unfriendly platforms like AIX are useful in some respects.

> Large thread stack size (8 mb) may help.

Increasing your stack size to 8mb is almost certainly masking some
real problem somewhere.

>      and what compile / load options are appropriate for Forte compiler?

Your mail mentions -D_REENTRANT for the Forte compiler but you really
need "-mt" instead.

> Purify on Solaris
> found no problems in 2 runs.

Purify claims to detect stack overflows but in practice this feature
rarely works for me.

> AIX: -lpthread -qthreaded

How are you invoking the compiler?  "xlC_r .... -lpthread -qthreaded"


> Ideas:  convert fork() & execle() to
>             system()
>             on AIX
>
>             Devote 1 thread to fork & exec
>
>             Find purify-like compile + run options for AIX.

If it were me, I would seriously consider porting this program over to
g++ (possibly on Solaris).  Then, I would compile the program with
full warnings turned on (-w, -pedantic), etc. and I would fix all of
the problems.  Then I would compile the program with -fstack-check
(note: in a multi-threaded program some code changes might be needed
here).  I would also turn on all warnings with the Solaris and AIX
compilers and fix these problems too.

--kevin
-- 
GnuPG ID: B280F24E                     And the madness of the crowd
alumni.unh.edu!kdc                     Is an epileptic fit
                                       -- Tom Waits