[GSoC2016] Weekly Report of Porting musl to RISC-V Project #4

[GSoC2016] Weekly Report of Porting musl to RISC-V Project #4

[Masanori Ogino]

Hello all,

This is the fourth issue of weekly report on porting musl to RISC-V
GSoC project.

This week, I have tried to run libc-test [1] on Spike with the
musl-based toolchain, but not succeeded. I found a minor problem with
recent cpp (I will report it to the maintainer of libc-test later) but
it is not the main issue preventing the test. I will continue
debugging this weekend.

There is no significant progress on the project this week. I will try
to contact other developers more and more to gain speed.

Have a nice weekend!

[1]: http://nsz.repo.hu/git/?p=libc-test
-- 
Masanori Ogino

Re: [GSoC2016] Weekly Report of Porting musl to RISC-V Project #4

[Szabolcs Nagy]

* Masanori Ogino <masanori.ogino@gmail.com> [2016-06-18 22:21:58 +0900]:
> This week, I have tried to run libc-test [1] on Spike with the
> musl-based toolchain, but not succeeded. I found a minor problem with
> recent cpp (I will report it to the maintainer of libc-test later) but

you have to s/-E/-E -P/ in the makefile

i was reluctant to do this upstream as it is a non-standard flag.

Re: [GSoC2016] Weekly Report of Porting musl to RISC-V Project #4

[Masanori Ogino]

2016-06-19 1:30 GMT+09:00 Szabolcs Nagy <nsz@port70.net>:
> * Masanori Ogino <masanori.ogino@gmail.com> [2016-06-18 22:21:58 +0900]:
>> This week, I have tried to run libc-test [1] on Spike with the
>> musl-based toolchain, but not succeeded. I found a minor problem with
>> recent cpp (I will report it to the maintainer of libc-test later) but
>
> you have to s/-E/-E -P/ in the makefile
>
> i was reluctant to do this upstream as it is a non-standard flag.

Thank you for letting me know!

-- 
Masanori Ogino

Re: [GSoC2016] Weekly Report of Porting musl to RISC-V Project #4

[Rich Felker]

On Sat, Jun 18, 2016 at 06:30:17PM +0200, Szabolcs Nagy wrote:
> * Masanori Ogino <masanori.ogino@gmail.com> [2016-06-18 22:21:58 +0900]:
> > This week, I have tried to run libc-test [1] on Spike with the
> > musl-based toolchain, but not succeeded. I found a minor problem with
> > recent cpp (I will report it to the maintainer of libc-test later) but
> 
> you have to s/-E/-E -P/ in the makefile
> 
> i was reluctant to do this upstream as it is a non-standard flag.

If gcc isn't planning to fix this, I think we should find a good
out-of-the-box workaround in libc-test. That would probably be either
a probe for gcc needing -P, or a filter to undo whatever mangling gcc
does without -P. It seems like a fairly trivial sed script could do
the latter.

Rich

Re: [GSoC2016] Weekly Report of Porting musl to RISC-V Project #4

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 894 bytes --]

On Jun 26, 2016 9:12 AM, "Rich Felker" <dalias@libc.org> wrote:
>
> On Sat, Jun 18, 2016 at 06:30:17PM +0200, Szabolcs Nagy wrote:
> > * Masanori Ogino <masanori.ogino@gmail.com> [2016-06-18 22:21:58 +0900]:
> > > This week, I have tried to run libc-test [1] on Spike with the
> > > musl-based toolchain, but not succeeded. I found a minor problem with
> > > recent cpp (I will report it to the maintainer of libc-test later) but
> >
> > you have to s/-E/-E -P/ in the makefile
> >
> > i was reluctant to do this upstream as it is a non-standard flag.

Just pin minimum gcc version to 5

>
> If gcc isn't planning to fix this, I think we should find a good
> out-of-the-box workaround in libc-test. That would probably be either
> a probe for gcc needing -P, or a filter to undo whatever mangling gcc
> does without -P. It seems like a fairly trivial sed script could do
> the latter.
>
> Rich

[-- Attachment #2: Type: text/html, Size: 1229 bytes --]

Re: [GSoC2016] Weekly Report of Porting musl to RISC-V Project #4

[Rich Felker]

On Sun, Jun 26, 2016 at 03:51:50PM -0700, Khem Raj wrote:
> On Jun 26, 2016 9:12 AM, "Rich Felker" <dalias@libc.org> wrote:
> >
> > On Sat, Jun 18, 2016 at 06:30:17PM +0200, Szabolcs Nagy wrote:
> > > * Masanori Ogino <masanori.ogino@gmail.com> [2016-06-18 22:21:58 +0900]:
> > > > This week, I have tried to run libc-test [1] on Spike with the
> > > > musl-based toolchain, but not succeeded. I found a minor problem with
> > > > recent cpp (I will report it to the maintainer of libc-test later) but
> > >
> > > you have to s/-E/-E -P/ in the makefile
> > >
> > > i was reluctant to do this upstream as it is a non-standard flag.
> 
> Just pin minimum gcc version to 5

libc-test does not depend on gcc much less a particular gcc version.
Ability to apply tests with different compilers is fairly important
since, for some interfaces, a buggy compiler might transform how
they're invoked in invalid ways that result in breakage, and these
sorts of bugs are useful to be able to track down.

Rich