tech-pkg archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: wrappers and gcc -isystem



Jonathan Perkin <jperkin%joyent.com@localhost> writes:

> * On 2016-02-18 at 12:34 GMT, Greg Troxel wrote:
>
>> Jonathan Perkin <jperkin%joyent.com@localhost> writes:
>> 
>> >> [1] What's keeping this, anyway?
>> >
>> > I don't know but I'd like to see it happen 12 months ago.  The
>> > difference in reduced build/system time is astonishing and I've used
>> > it for all my builds for a long time now.
>> 
>> Perhaps you could start a new thread with the accumulated evidence that
>> enabling cwrappers does no harm, in terms of missing packages in bulk
>> builds, and packages that can't be build normally, on various platforms.
>> Because I haven't seen a "here is the data; I think we're over the bar",
>> I have been under the impression that we hadn't yet arrived at time to
>> flip the default.
>
> This really needs to be done on NetBSD, as the only platform which can
> currently build almost all of pkgsrc.  I don't have any NetBSD bulk
> build infrastructure, so this will need to come from someone else.

So we need Someone to run a bulk build on NetBSD with and without.

> However, for the bulk builds I perform on SmartOS, OSX and Linux, I'm
> not aware of any regressions caused by moving to cwrappers, just
> significantly faster builds.

That sounds like "no missing packages, compared to building w/o
cwrappers".

>> Also, it seems one probably has to re-bootstrap.  I am unclear on what
>> the effect on NetBSD itself is and how that would happen, in terms of
>> changing the default and getting existing systems moved over.
>
> This isn't the case.  Just add USE_CWRAPPERS=yes to mk.conf and you're
> good to go.

That makes it easier.  I have two netbsd-7 machines (i386, amd64) that I
do rebuilds on (pkg_rr, not bulk) to look for packages that need
fixing.   After they finish the current round I'll add that.

> I did add a '--cwrappers' argument to the bootstrap script to make it
> easier to test, perhaps that is causing the confusion.  In hindsight I
> probably shouldn't have added it.

I didn't know about that.   It was just not obvious to me whether this
was a bootstrap thing or a runtime thing, without looking.

Attachment: signature.asc
Description: PGP signature



Home | Main Index | Thread Index | Old Index