r191394 - Produce an error for unknown -f options.
rnk at google.com
Thu Sep 26 08:50:13 PDT 2013
On Thu, Sep 26, 2013 at 6:16 AM, Rafael Espíndola <
rafael.espindola at gmail.com> wrote:
> I implemented most of them in r191429.
> I left out:
> * -fplugin-*: clang plugins are different from gcc. We should probably
> err on someone trying to pass us a gcc plugin.
> * -fbuiltin-*: I could not find an easy way to get the list that gcc
This can be done with a joined option:
def fbuiltin_ : Joined<["-"], "fbuiltin-">;
def fno_builtin_ : Joined<["-"], "fno-builtin-">;
Which Nick says is how we would actually implement it now that LLVM has TLI.
> * options that gcc itself errs on (like the -fwarn-incomplete-patterns)
> * options we already implemented (like -fasynchronous-unwind-tables)
> Were you grepping error logs?
> On 25 September 2013 22:07, Nick Lewycky <nlewycky at google.com> wrote:
> > On 25 September 2013 18:53, Rafael Espíndola <rafael.espindola at gmail.com
> > wrote:
> >> >> Err, we aren't ready for this. If we're going to do this, could you
> >> >> stage
> >> >> it such that the gcc flags are added but ignored, then we error on
> >> >> unknown
> >> >> options?
> >> >
> >> >
> >> > This is already how it works. If our list of ignored gcc -f flags
> >> > complete enough, that's simple to fix.
> >> I am happy to revert if needed, but as Eli pointed out adding ignored
> >> options is really easy. Do you have a list?
> > Using regex syntax:
> > -falign-functions.*
> > -fasynchronous-unwind-tables
> > -fbuiltin-.*
> > -fcallgraph-profiles-sections
> > -ffloat-store
> > -feliminate-unused-debug-types
> > -fgcse
> > -fident
> > -fivopts
> > -fnon-call-exceptions
> > -fpermissive
> > -fplugin-arg-.*
> > -fplugin=.*
> > -fprefetch-loop-arrays
> > -fprofile-correction
> > -fprofile-dir.*
> > -fprofile-use.*
> > -fprofile-values
> > -frounding-math
> > -fripa
> > -fsee
> > -ftracer
> > -ftree-.*
> > -funroll-all-loops
> > -fwarn-incomplete-patterns
> > and the -fno- versions of each of these. I may also be missing more, but
> > this is a start.
> > Nick
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cfe-commits