Removing the glibc NaCl port

Removing the glibc NaCl port

[Florian Weimer]

I just realized that my pipe2 removal probably breaks the NaCl port.

Can we remove it?  It blocks useful cleanups which are compatible with 
the out-of-tree Hurd port.

I don't think anyone has ever reported test results for release builds. 
scripts/build-many-glibcs.py lacks support for it, so we don't even know 
if it still builds.

I hope this proposal is not too controversial.

Thanks,
Florian

Re: Removing the glibc NaCl port

[Adhemerval Zanella]

On 13/04/2017 12:45, Florian Weimer wrote:
> I just realized that my pipe2 removal probably breaks the NaCl port.
> 
> Can we remove it?  It blocks useful cleanups which are compatible with the out-of-tree Hurd port.
> 
> I don't think anyone has ever reported test results for release builds. scripts/build-many-glibcs.py lacks support for it, so we don't even know if it still builds.
> 
> I hope this proposal is not too controversial.
> 
> Thanks,
> Florian

Also last time I check from official documentation [1], it uses ancient
version without any documentation to actually build/check glibc on it.
I have the impression it is bit-rotten for some time.

If the idea is to keep supporting, I think we need at least proper
documentation on how to actually build and tests, preferable with 
build-many-glibc.py support. 

[1] https://developer.chrome.com/native-client/sdk/download

Re: Removing the glibc NaCl port

[Florian Weimer]

On 04/13/2017 07:44 PM, Adhemerval Zanella wrote:
> 
> 
> On 13/04/2017 12:45, Florian Weimer wrote:
>> I just realized that my pipe2 removal probably breaks the NaCl port.
>>
>> Can we remove it?  It blocks useful cleanups which are compatible with the out-of-tree Hurd port.
>>
>> I don't think anyone has ever reported test results for release builds. scripts/build-many-glibcs.py lacks support for it, so we don't even know if it still builds.
>>
>> I hope this proposal is not too controversial.


> Also last time I check from official documentation [1], it uses ancient
> version without any documentation to actually build/check glibc on it.
> I have the impression it is bit-rotten for some time.
> 
> If the idea is to keep supporting, I think we need at least proper
> documentation on how to actually build and tests, preferable with
> build-many-glibc.py support.
> 
> [1] https://developer.chrome.com/native-client/sdk/download

There is a Git repository with a NaCl port to GCC 4.9, it seems:

https://chromium.googlesource.com/native_client/nacl-gcc/+/ng/4.9/master

Version-wise, this is sufficient to build glibc (unlike the 
GCC-4.4-derived version in Fedora).  I have not checked if it actually 
works.  GCC upstream no longer supports any of the 4.x branches, so this 
is a dead end anyway.

The NaCl SDK seems to have switched to Clang/LLVM, and therefore cannot 
build glibc anymore.  (I checked the pepper_56 bundle.)

Thanks,
Florian