[webkit-dev] Build slave for JSCOnly Linux MIPS

[Filip Pizlo]

> On Apr 19, 2016, at 10:11 AM, Michael Catanzaro <mcatanzaro at igalia.com> wrote:
> 
> On Mon, 2016-04-18 at 17:26 -0700, Filip Pizlo wrote:
>> Yeah.  If we allow GCC 4.8 then I think we should make all of our
>> code compile with it.  If that proves too difficult (like if we had
>> to get rid of all lambdas), then I think we need to not allow GCC
>> 4.8.
> 
> Agreed. And don't worry, we would not have to get rid of lambdas...
> that would not be a reasonable proposal.
> 
> I can confirm that WebKitGTK+ 2.10.7 works fine with GCC 4.8.5 when
> patched to remove the error we added to require use of GCC 4.9, and
> when using -DENABLE_DATABASE_PROCESS=OFF and -DENABLE_INDEXED_DB=OFF
> (undesirable but acceptable IMO). I'm not sure if the database process
> disablement is still required with the latest GCC 4.8 release, but it
> was crashing at some point in the past.

Crashing?  Ouch!

I think we should get to the bottom of this.  If we can lower the requirement to GCC 4.8.5 with those DB flags on, then this strengthens the argument for lowering the GCC version requirement.  On the other hand if that process crashes when compiled with GCC 4.8.5 then the risk is that there is still a serious compiler bug that affects our style of programming and this bug will be revealed in other places once someone breathes on the code the wrong way.

Like any big project, we have done hacks to support buggy compilers and I think that this is often the pragmatic thing to do.  I’m starting to think that we should probably do this for GCC 4.8.5, but I’d like some more insight into just how gross those hacks are.

> 
> Now, there are eight months of changes in trunk since that last
> confirmed-working version; if we were to lower the supported GCC
> version, we would have to investigate to see if it still works with
> trunk.
> 
> Michael