This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [patch,fortran] Fix corner case where consecutive calls to date_and_time can return times out of sequence (PR30015)


Hi Francois-Xavier,


FX Coudert wrote:
>> In addition, I removed the configure check whether gettimeofday supports
>> a timezone or not. The timezone was never used.
>> (gettimeofday with two arguments is defined in POSIX, a non-NULL second
>> argument is undefined in POSIX.) The removal of this AC_TRY_RUN
>> increases the cross-compilation support.
>
> Also true. But unless you have a reason for also removing
> GETTIMEOFDAY_ONE_ARGUMENT, I think it would be better to keep it in
> acinclude.m4, and use it in date_and_time.c.

Well, there is:

1. To my knowledge all GCC platforms offer either no gettimeofday (not
even with one argument) or a two-argument gettimeofday; besides
libgfortran, gettimeofday is currently used by
gcc/toplev.c, include/libiberty.h, libiberty/mkstemps.c,
libiberty/gettimeofday.c,
libjava/classpath/native/jni/java-lang/java_lang_VMSystem.c,
libmudflap/mf-runtime.c
and none of them uses a one-argument version of gettimeofday.

2. As Andreas Schwab wrote, removing this configure test "would also
make it possible to eliminate a runtime configure check." [...] "Runtime
checks are bad for cross compiling."

But if you really think that having a run-time check for
GETTIMEOFDAY_ONE_ARGUMENT outweighs the problems of run-time tests/cross
compiling (and of having more code ;-), I can re-add the check.

Tobias


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]