ml_utils.c: change use of c++11 posix standard function "gethostname"…

[prwolfe]

… to c99 standard "uname"

Since cmake interprets this as a c file by name it
only pulls in posix functions from the 98 standard.
That meant that "gethostname" was an implicit declaration
and that fails when running with warnings-as-errors.

[aprokop]

@trilinos/ml

[ibaned]

I don't think these functions have anything to do with the C++11 and C99 standards at all? They are both in the POSIX.1-2001 standard. POSIX is only a C API standard, has nothing to do with C++, and as far as I know also doesn't depend on the version of the C standard being used. Can you post the configuration you were using and the warning that was emitted?

[prwolfe]

Sure - this comes directly from the dashboard in the sierra replication builds. See https://testing.sandia.gov/cdash/viewBuildError.php?buildid=3143344

I did look into the unistd.h header and the notes from the gcc compiler team are what I was referencing as to the versions. I did confirm that the macros in units.h turn off declaration of gethostname when using -std=c99 which this does. I also tried to use -std=c++11 and the compiler said no as it's c code. We could work through and get it to use -std=c11 (we currently don't pass anything - not sure where c99 came from.)

[prwolfe]

It looks like the compiler default will be c11 with gnu extensions. I think I will talk with Jim and Mike to see if all of Trilinos should be using that or the stricter c11. Sierra does not pass anything and so would get the gnu extensions.

Thanks!

[ibaned]

Hmm you're right that its not being declared properly, I just confirmed while building Trilinos:

/home/daibane/src/Trilinos/packages/ml/src/Utils/ml_utils.c: In function ‘ML_BreakForDebugger’:
/home/daibane/src/Trilinos/packages/ml/src/Utils/ml_utils.c:1965:9: warning: implicit declaration of function ‘gethostname’ [-Wimplicit-function-declaration]
         gethostname(hostname, sizeof(hostname));
         ^~~~~~~~~~~

[prwolfe]

@jwillenbring - did this ever get discussed?

[mhoemmen]

@prwolfe Not that I recall, though I'm not @jwillenbring :-) . I would say, if it works, merge it.

[prwolfe]

Pull request #1974 will take care of this until a full policy is established.

[prwolfe]

Closed and moved to issue #1985