GCC 4.7.2 : Get it into DragonFly 3.2 or wait for 3.3 branch?

John Marino dragonflybsd at marino.st
Mon Oct 1 07:02:27 PDT 2012


On 9/28/2012 04:31, Justin Sherrill wrote:
> On Tue, Sep 25, 2012 at 3:27 AM, John Marino<dragonflybsd at marino.st>  wrote:
>
>> Assuming all the changes necessary to bring in GCC 4.7.2 are complete by
>> Sunday, I'd recommend waiting 8 days (the following Monday) to branch and
>> maintaining the 14 days to release.  I think 8 days would be enough time to
>> flush out any world build problems, not that I plan on introducing any.
>>
>> It also gives a little more time to flush out any anomalies with the
>> revamped usched_dfly.
>
> I am OK with this - anyone else object?
>
> John, do you have time for those changes?

Hi Justin,
I ran into i386-specific and bootstrapping issues that took a couple of 
days to work though.  I have committed gcc47 to the trunk though. 
Here's some things to note:

1) GCC44 can build world include GCC47 just fine
2) GCC47 can't build world yet because it has a few new error messages 
that prevent the build from completing when -Werror is used.
3) Theoretically setting NO_WERROR globally would work around this, but 
I haven't done it yet
4) I'm slowly fixing all the codes issues with WORLD_CCVER=gcc47 but 
several breaks remain
5) There's an incompatibility issue with gcc4.4-built CSU files that 
results in no error handling frame getting added when WORLD_CCVER=gcc47. 
  I need to work through that.

The bottom line is people can use gcc47 through the CCVER=gcc47 
mechanism, but nobody should expect to use WORLD_CCVER=gcc47 
successfully yet.  That's forward work.  Hopefully we'll be further 
along before the branch.

GCC41 has been disabled, so only two compilers will build on trunk.

Regards,
John



More information about the Users mailing list