This is the mail archive of the gcc@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: multilib build problem with default options


On Tue, 2012-06-19 at 20:59 +0000, Joseph S. Myers wrote:
> On Tue, 19 Jun 2012, Steve Ellcey  wrote:
> 
> > I am building a multilib mips compiler using '--with-synci' on the 
> > configure line.  This means that gcc will pass the '-msynci' option
> > to cc1 by default.  The problem occurs when one of my multilib
> > targets (like -mips32) does not support synci.  This means that
> > any compilation done with -mips32 will generate a warning during
> > compilation.  Unfortunately, this means that I have configure 
> 
> I think you need to fix this problem rather than trying more and more 
> complicated workarounds.  Essentially, the spec in OPTION_DEFAULT_SPECS 
> can't simply be "%{!msynci:%{!mno-synci:-m%(VALUE)}}" - it needs to be 
> something more complicated that avoids passing -msynci when it won't be 
> supported, or else it needs to pass an internal -msynci-if-supported 
> option which acts like -msynci except for not causing a warning.

Trying to figure out when an architecture won't support -msynci seems
pretty complicated due to the many ways of specifying the various
flavors of mips so I am not sure I could write a SPEC to do that
correctly, but passing an internal -msynci-if-supported flag instead of
passing -msynci seems like a pretty straight forward way to get the
desired behavior.  I will pursue that option.  Thanks for the advice.

Steve Ellcey
sellcey@mips.com


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