[BUG SUB] silc toolkit contains gcc-isims and CFLAGS when it shouldn\'t - silc toolkit

[Torrey McMahon]

Pekka Riikonen wrote:
> On Sun, 3 Jun 2007, Torrey McMahon wrote:
>
> : > compiler/OS you were using (what configure said you were using) so that we
> : > can set specific case for that.
> : 
> : Solaris and Sun's compilers. You'd want to set -xO[N] where N is between 
> : 1 and 5. However, with the Sun compilers the options are parsed first to 
> : last so if I set my CFLAGS to max optimization and then the configure 
> : script comes in and decides to put "-g -xO2" at the end it overrides 
> : everything I just set.
> : 
> Yes, well, that can be considered a bug.  The order indeed could be so 
> that what user gives always stays first.  I shall fix this.  Are there no 
> -O options with the compiler and if there are, don't they do the same 
> thing the -xOn does?
>   

In my case the user flags need to be last otherwise the addition of "-g 
-O" or even "-g -xO2" removes the heavier optimizations I've already set 
in CFLAGS. Why not have a default set of CFLAGS but if the user has 
already set one then the users flags are taken without additions or 
modifications?


> : > : Second, MATH_CFLAGS is set and used in some locations and its gcc : 
> : > specific. That should either be an option or env variable to set at : 
> : > configure time. > : That's one of the parts we want to compile and 
> : optimize differently, hence > the MATH_CFLAGS.  Even if I'd add support 
> : for MATH_CFLAGS environment > variable we'd still set our default, 
> : mandatory, options for all compilers.
> : 
> : The options you're using are gcc dependent and cause lots of error messages
> : when using Sun's compiler.
> : 
> That's weird.  All options are first checked that they work with the 
> compiler.  So whatever option it puts there the compiler obviously have 
> accepted.  That of course doesn't mean the options are correct options.
>   

The compiler outputs errors. It doesn't stop working but I get lots of 
output about how fsimple and the like aren't flags to pass to the compiler.

> Could you please tell me what platform ./configure said you were using.  
> And if you have figured out with what options the math library works could 
> you tell me what they are.

Solaris i386 using the Sun compilers and not gcc. I'm not sure what the 
gcc flags you've set do but I'd perform a check that says, "If we're not 
using gcc then don't set MATH_CFLAGS".