Proposal for the merger of devel/modula-2 into master

Proposal for the merger of devel/modula-2 into master

[Gaius Mulley]

Hi David,

David Edelsohn <dje.gcc@gmail.com> writes:

> I hope that you and the GNU Modula-2 team can propose the merge of the
> Modula-2 front-end and library soon.

[reposting with a new title for mailing list clarity]

Yes I was also wondering whether now might be a sensible time for the
Modula-2 front end to move across onto the master branch?

The current Modula-2 front end and libraries are in the branch:
devel/modula-2 and it builds on gnu/linux/amd64, gnu/linux/aarch64,
gnu/linux/sparc64, solaris/sparc64, gnu/linux/i386, gnu/linux/ppc64le,
gnu/linux/power7 (and probably others).  (Test results posted to
gcc-testresults).

More than happy for the gm2 front end to be merged and/or help or
receive instructions on how to progress the merge

regards,
Gaius

Re: Proposal for the merger of devel/modula-2 into master

[Richard Biener]

> Am 14.05.2022 um 00:57 schrieb Gaius Mulley via Gcc <gcc@gcc.gnu.org>:
> 
> 
> Hi David,
> 
> David Edelsohn <dje.gcc@gmail.com> writes:
> 
>> I hope that you and the GNU Modula-2 team can propose the merge of the
>> Modula-2 front-end and library soon.
> 
> [reposting with a new title for mailing list clarity]
> 
> Yes I was also wondering whether now might be a sensible time for the
> Modula-2 front end to move across onto the master branch?

Yes, it’s stage1 now.

> The current Modula-2 front end and libraries are in the branch:
> devel/modula-2 and it builds on gnu/linux/amd64, gnu/linux/aarch64,
> gnu/linux/sparc64, solaris/sparc64, gnu/linux/i386, gnu/linux/ppc64le,
> gnu/linux/power7 (and probably others).  (Test results posted to
> gcc-testresults).

There’s at least BSD in the list of primary targets. It would be nice to enable more but that can be done during stage1.

I suggest you post merge patches where the branch touches generic code for review again, indicating parts that have been approved in the past.

Richard 

> 
> More than happy for the gm2 front end to be merged and/or help or
> receive instructions on how to progress the merge
> 
> regards,
> Gaius

Re: Proposal for the merger of devel/modula-2 into master

[Gaius Mulley]

Richard Biener <richard.guenther@gmail.com> writes:

>> Am 14.05.2022 um 00:57 schrieb Gaius Mulley via Gcc <gcc@gcc.gnu.org>:
>> 
>> 
>> Hi David,
>> 
>> David Edelsohn <dje.gcc@gmail.com> writes:
>> 
>>> I hope that you and the GNU Modula-2 team can propose the merge of the
>>> Modula-2 front-end and library soon.
>> 
>> [reposting with a new title for mailing list clarity]
>> 
>> Yes I was also wondering whether now might be a sensible time for the
>> Modula-2 front end to move across onto the master branch?
>
> Yes, it’s stage1 now.
>
>> The current Modula-2 front end and libraries are in the branch:
>> devel/modula-2 and it builds on gnu/linux/amd64, gnu/linux/aarch64,
>> gnu/linux/sparc64, solaris/sparc64, gnu/linux/i386, gnu/linux/ppc64le,
>> gnu/linux/power7 (and probably others).  (Test results posted to
>> gcc-testresults).
>
> There’s at least BSD in the list of primary targets. It would be nice to enable more but that can be done during stage1.
>
> I suggest you post merge patches where the branch touches generic code for review again, indicating parts that have been approved in the past.
>
> Richard 

Hi Richard,

thanks for the email - sure I'll post the generic code patches with
references to those components which were okayed.  I'm in the process of
building on BSD (Free and Open) platforms.

regards,
Gaius

Proposal for the merger and patch set

[Gaius Mulley]

Richard Biener <richard.guenther@gmail.com> writes:

> I suggest you post merge patches where the branch touches generic code
> for review again, indicating parts that have been approved in the
> past.

Hello Richard, David and the GCC Steering Committee,

Firstly thank you for the release of gcc-12.1 and secondly thank you
for taking the time to review the Modula-2 front end and for valuable
review feedback over the years.  I hope you have all managed to have a
break after the 12.1 release.

Here for reference are the previous patch emails and reviews

[For historical completeness
https://gcc.gnu.org/pipermail/gcc-patches/2019-November/534683.html
]

Jan 18 2021 patches:

https://gcc.gnu.org/pipermail/gcc-patches/2021-January/563732.html

Richard's <rguenther@suse.de> review of the patches last year.

https://gcc.gnu.org/pipermail/gcc-patches/2021-January/563747.html


June 2021 patches:

https://gcc.gnu.org/pipermail/gcc-patches/2021-June/573103.html

Richard's <rguenther@suse.de> review of the June patches last year.

https://gcc.gnu.org/pipermail/gcc-patches/2021-June/573147.html

Richard Guenther  <rguenther@suse.de> writes:

> I suggest you post merge patches where the branch touches generic
> code for review again, indicating parts that have been approved in
> the past.

Sure I'll try and group the subsequent email patches into the following categories:

  1.  all GCC files which have been patched.
  2.  the top level /gm2tools contents.
  3.  the top level /libgm2 contents.
  4.  the glue code (between Modula-2 and GCC) contained in /gcc/m2/gm2-gcc.
      (these have not been reviewed before).  These will be posted in
      3 sub parts.
  5.  all the driver code /gcc/m2.
  6.  /gcc/m2/Make-lang.in
      (not been reviewed before)
  7.  /gcc/doc/gm2.texi
      (not been reviewed before)

The full source code is available via:

  $ git clone git://gcc.gnu.org/git/gcc.git gcc-git-devel-modula2
  $ cd gcc-git-devel-modula2
  $ git checkout devel/modula-2

currently it is being merged every 1-3 days or so.  One of the
recommendations after the last review was to maintain this branch.
Over the last year the changes include many bug porting fixes, removal
of compiler build warnings, migrating the code base to C++, improving
accuracy/detail of compiler error messages and improving error recovery.

It builds on gnu/linux/amd64, gnu/linux/aarch64, gnu/linux/sparc64,
solaris/sparc64, gnu/linux/i386, gnu/linux/ppc64le, gnu/linux/power7
(and probably others).  Also built with profiled lto on amd64 and
aarch64.

Richard Guenther  <rguenther@suse.de> writes

> There’s at least BSD in the list of primary targets. It would be
> nice to enable more but that can be done during stage1.

and now FreeBSD 13 i586 providing a patch was applied for C++.
[Interestingly csinl, ctanl and friends are missing on FreeBSD 13
which cause all runtime tests to fail].

(Test results posted to gcc-testresults).

On amd64-gnu-linux it builds as a cross compiler for aarch64.
http://floppsie.comp.glam.ac.uk/Southwales/gaius/web/bare-metal-m2.html

I've built GCC from the master branch enabled c,c++,fortran,go,d
and then built GCC from devel/modula-2 enabled c,c++,fortran,go,d,m2
and no further regressions were seen.

Anyway I suspect this email is long enough for now - feel free to ask
questions, 9 patches posted to the gcc-patches list.  Let me know if you
require the patch set and/or new files in a different format,

regards,
Gaius