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: MPX runtime inclusion for GCC 5

2015-02-26 11:30 GMT+03:00 Richard Biener <>:
> On Wed, 25 Feb 2015, Jeff Law wrote:
>> So the steering committee has agreed to include the MPX runtime in GCC.  The
>> runtime bits are a necessary component to actually make use of the MPX
>> capabilities.
>> GCC is a downstream consumer of the core runtime and the canonical repository
>> for the wrapper subpart of that runtime.
>> The question is whether or not to try and get the MPX runtime integrated for
>> GCC 5 or wait until the next stage 1.
>> I'm hesitant to add at this stage, but if we could limit it to only building
>> on host/target combinations where we know it works, then perhaps including it
>> might make sense as it'll be easier for developers who want to take advantage
>> of MPX to get everything they need from the GCC release rather than stitching
>> things together from multiple locations.
>> Jakub, Joseph & Richi have the final say here for inclusion in GCC 5.
> I'm undecided here - if we can make sure to not break anybodies bootstrap
> for example by either disabling the build of the MPX runtime by default
> (thus make a "virtual" --with-system-mpx-runtime the default) or by
> only enabling it by default for known targets (but as seen with the
> sanitizers things break even then with dependencies on glibc or kernel
> header details) then I'm fine with it.  After all the MPX feature
> is "incomplete" right now?  I woudln't even know where to get the
> MPX runtime from (I suppose it's not documented anywhere in install.texi
> nor is it placed in infrastructure/ or supported to be built in-tree).

You are right that without MPX runtime libraries feature is incomplete
and harder to use. Users would have to download, build, install and
link them separately.

In current patches libmpx is enabled by default only for
x86_64-*-linux* and i?86-*-linux*. I agree that current configure may
miss some system checks. With amount of time left I think having it
disabled by default is an appropriate option. It should be switched
back on stage 1 and/or minor releases.


> Richard.
> --
> Richard Biener <>
> SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Jennifer Guild,
> Dilip Upmanyu, Graham Norton HRB 21284 (AG Nuernberg)

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