This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATH: call move init_reg_modes call (for compile-server)
Zack Weinberg wrote:
That's understandable, but I would be more comfortable with the change
if you did the investigation to find out what exactly is going on here.
I think is this an incremental process. As soon as the compile server
branch goes active (delayed because I have to resolve various merge
conlficts first), and we agreed on a naming convention, then we can
look at the various init functions and figure out where the go.
For example, I glanced at init_eh, and it seems like most or all of
that falls into the _once category. But note it needs to be called
after the front-end language_init_once function.
We get a potentially further layer of initialization if we allow
the compile server to change command-line flags between "runs".
In that case some of the _once functions may have to be re-done,
and they are no longer _once. However, I'm hoping we will be
able to handle this by dependency checking and the same invalidation
that happens when include files have to be re-read, so hopefully
this won't be a big issue.
--
--Per Bothner
per@bothner.com http://per.bothner.com/