This is the mail archive of the
mailing list for the GCC project.
Re: Patch to fix -fsyntax-only
- To: bothner at cygnus dot com
- Subject: Re: Patch to fix -fsyntax-only
- From: craig at jcb-sc dot com
- Date: 19 Feb 1999 15:20:09 -0000
- Cc: egcs-patches at egcs dot cygnus dot com
- Cc: craig at jcb-sc dot com
- References: <199902182225.OAA09696@cygnus.com>
>> Java people -- I take it you have a specifical functional need
>> for -fsyntax-only working properly, and probably make use of
>> areas of the front end that g77 doesn't.
>Yes - when compiling Java source to bytecode, the gcj front-end
>passes the flags -fsyntax-only -femit-class-files to jc1.
>So, from our point of view, -fsyntax-only should suppress any
>generation or RTL or assembly code, but not suppress generation
>of tree nodes, error checking, or semantic analysis.
Do you have a *functional* need to suppress generation of the RTL
Reason I ask is, currently, RTL *is* generated, unless you do all
your code generation exclusively via expand_expr_stmt (meaning you
don't call expand_jump, expand_start_cond, etc.).
What -fsyntax-only does inhibit is most or all of the expensive
RTL transformations -- i.e. optimizations.
Please read the longish email I sent Jeff, on this same subject,
a few minutes ago, for more information.
tq vm, (burley)
P.S. In the meantime, I'm doing further testing on my patch, and
am trying to figure out a strange behavior that might have nothing
to do with it. But it might, so don't do anything with my patch