| Re: [AD] gcc 3.x workaround. |
[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
George Foot <gfoot@xxxxxxxxxx> wrote:
> What about a mmxtest-like solution for platforms that don't
> support configure scripts? It could be as simple as testing the
> gcc version and warning the user if their version required that
> switch to be turned off; alternatively, the program could modify
> an included makefile and rerun make automatically.
The problem with this is that it is a messy solution on the dos/win32
platform. It's far from as elegant as it could be done on a unix shell.
On mingw one can make one test program, include libiberty.h, and check the
preprocessor define GCC_VERSION.
Alternatively one can rely on sed and check the output from gcc -v or
gcc --version. mingw doesn't have sed in it's current toolchain FYI.
Putting back -fomit-frame-pointer is IMO a bad idea. Doing so would at least
require the docs to be very clear about this problem. (it should be
mentioned in every relevant readme) and even if that is done be sure to
receive many unneccessary bugreports on the mailinglist. :o) This is only a
solution if profiling proves -fomit-frame-pointer gives a significant speed
advantage. To be honest I haven't tested this yet...
Just removing -fomit-frame-pointer is simple. I wouldn't mind another
solution if somebody tells me why it's better.
Sincerely Henrik Stokseth.
-----------------------------------------------------------------------
E-mail: hstokset@xxxxxxxxxx Homepage: http://hstokset.n3.net