BAIHELP%TAMSTAR.BITNET@MITVMA.MIT.EDU (07/20/89)
Well --
Having read the letter from Chris Siebenmann about how Unix normally
ignores a return value if one is not supplied, I still has a problem.
Isn't it the same thing to want to return a "all ok" signal to the
operating system if no other value is supplied (ie. exit(-1);).
I assume that Unix does return "some kind" of value if one is not
supplied. This is what VMS is expecting. If you don't exit in a
particular way (ie. exit(-1);) then VAX C assumes everything went ok,
puts a 1 in r0 and returns.
Where should the code be modified to cause a return from main to be
different from a return from any other routine, in that a assumed
return value of 1 is supplied in r0.
Bryan A. Ignatow
VMS Software Specialist
Public Domain Software Specialist
Academic Vax Cluster
Texas A&M Univeristy
Bitnet: BAIHELP@TAMVENUS
Internet: BAIHELP@venus.tamu.eduBAIHELP%TAMSTAR.BITNET@MITVMA.MIT.EDU (07/21/89)
In reference to the problem of VMS 5.1 needing a longword 1 in r0 if
no other value is supplied:
o Unix lets you drop out of the bottom of main.
o Microsoft C lets you drop out of the bottom of main.
o From "The C Programming Language" by K&R pg. 26
A function need not return a value; a RETURN statement
with no expression causes control, but no useful value,
to be returned to the caller, as does "falling off the
end" of a function by reaching the terminating right
brace. And the calling function can ignore a value
returned by a function.
... Since MAIN is a function like any other, it may
return a value to its caller, which is in effect the
environment in which the program was executed.
... we have omitted RETURN statements from our MAIN
functions up to this point, but we will include them
hereafter, as a reminder that PROGRAMS SHOULD RETURN
STATUS TO THEIR ENVIRONMENT.
Hence, since I won't get into an arguement about "good programming
style", I will say that the problem lies not with the VAX C compiler
but with the operating system. VMS expects a status return value in
general purpose register 0 and if not supplied retruns an error the
the process running the program. Now if "in the interests of good
style" a programmer would use exit(n); to halt his program on error,
then sould we not assume that if a program is about to drop off the
end of main and hasn't set a prefered return value that one may be
assumed for it. This allows COMPATIBILITY with the VMS 5.1 operating
system.
I am not suggesting this change be made for all implementations of
GCC, only for those running under VMS which need it.
Bryan A. Ignatow
VMS Software Specialist
Public Domain Software Specialist
Academic Vax Systems
Texas A&M University
Bitnet: BAIHELP@TAMVENUS
Internet: BAIHELP@venus.tamu.edurms@mipos3.intel.com (07/21/89)
I would not mind installing changes for main to return 1 by default on VMS, if somebody else implements them cleanly.
msnyder@tut.cis.ohio-state.edu (Michael V Snyder) (08/02/89)
Brian and other VMS-using friends:
Instead of trying to make the compiler recognise main() as a
special case, would it not be easier to make EVERY function return
a 1 in R0 in case an explicit return is not performed? I don't think
this behavior should cause problems, as the behavior under these
circumstances is undefined. Any program that depends on a function
returning the value of it's last expression, say, probably shouldn't.
Michael Snyder
msnyder@csi.compuserve.com