This is the mail archive of the 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: [PATCH 1/2] documentation fixes #1

On 8/30/2010 8:59 AM, Richard Guenther wrote:


And "an LTO".

I'll fix the grammar. There was also another mistake noone noticed except me on re-reading :)

+Configure the compiler with @samp{--enable-stage1-languages=c,lto}. This
+will only work with the @command{gold} linker and if linker plugins are
+corrected configured.  The @option{-frandom-seed=1} is needed to avoid
+bootstrap comparison failures.
This sounds a bit tricky for our general user installation instructions;
any chance to have all this triggered by a configure option or a simple
"make bootstrap-lto" or just document it in pure developer documentation?
(More a question to Diego et al, I guess.)  Oh wait, that's what's there,
but broken, right? :-(
The -frandom-seed requirement is indeed very unfortunate.  Andi - can't we
use sth based on a hash instead that would be stable across bootstrap?

In theory could use a hash over the global symbol names, but I'm a bit wary of collisions.
For example if you have a set of machine generated files with very similar symbols names
there might be problems. Also the random seed issue is only really a problem for bootstraps,
but not for other uses of the compiler.

Otherwise I'd like us to get rid of --enable-intermodule and its makefile
bits and simply fix --with-build-config=bootstrap-lto.

I'm not very good with Makefiles so I would prefer if someone else does that.


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