Date: Thu, 16 Jul 2026 00:07:34 -0400
Thank you, Matthias! Sorry for the delayed response. All suggestions
applied with some editorial discretion; details below. Let me know if
further edits are desired.
On 7/13/26 10:09 PM, Matthias Wippich wrote:
> Hey,
> thanks for these excellent minutes and the Itanium ABI PR!
>
> A few small nits:
>> Matthias: I don't think we need to worry about other tools, they can invoke the preprocessor with the options needed.
> When I re-read this later, I'd like to be reminded that this was about
> the "preprocessing assembly sources with a C++ preprocessor" use case
> and the related language mode. I suggest
>> Matthias: I don't think we need to worry about other tools or the assembly-with-cpp language mode, they can invoke the preprocessor with the options needed.
Applied as suggested:
> Matthias: I don't think we need to worry about other tools or the
assembly-with-cpp language mode; users can invoke the preprocessor with
the options needed.
> Likewise, this is missing the second half of the point I made and has
> a small error in it:
>> Matthias: Linux .S sources depend on $ not being a digit, but it can be worked around with additional macro expansion.
> I suggest
>> Matthias: Linux .S sources depend on $ not being a nondigit, but it can be worked around with additional macro expansion. Linux does however also depend on $ being treated as nondigit for some of their .c sources.
Thank you! I recall being confused about that error when editing after
the meeting and assumed I misunderstood something.
Applied as:
> Matthias: Linux |.S| sources depend on |$| not being a nondigit, but
it can be worked around with additional macro expansion; however, some
of the Linux |.c| sources do depend on |$| being treated as nondigit.
> In the following quote
>> Aaron: I think there were some recently reported issues involving $ in the preprocessor. [...]
>> Matthias: One of those was reported by me.
> Only one of the links is to an issue (which was not reported by me).
> The other one is a PR by me, which is also linked in the paper - I
> believe I did point out that it's just implementation experience for
> Option 3 of the paper.
Applied as:
> Matthias: One of those is a PR submitted by me; it's linked in the
paper and is implementation experience for the proposed option 3.
Tom.
>
>
> Best regards,
> Matthias
>
> On Mon, Jul 13, 2026 at 5:37 AM Tom Honermann<tom_at_[hidden]> wrote:
>> The consistent behavior across compilers that use Itanium mangling inspired me to submit a PR to the Itanium ABI to add $ to the <identifier> non-terminal.
>>
>> https://github.com/itanium-cxx-abi/cxx-abi/pull/213: Allow '$' in identifiers
>>
>> Tom.
>>
>> On 7/12/26 8:06 PM, Tom Honermann via SG16 wrote:
>>
>> Minutes have been published for the 2026-07-08 SG16 meeting here. For those that attended, please review and offer corrections.
>>
>> Tom.
>>
>>
applied with some editorial discretion; details below. Let me know if
further edits are desired.
On 7/13/26 10:09 PM, Matthias Wippich wrote:
> Hey,
> thanks for these excellent minutes and the Itanium ABI PR!
>
> A few small nits:
>> Matthias: I don't think we need to worry about other tools, they can invoke the preprocessor with the options needed.
> When I re-read this later, I'd like to be reminded that this was about
> the "preprocessing assembly sources with a C++ preprocessor" use case
> and the related language mode. I suggest
>> Matthias: I don't think we need to worry about other tools or the assembly-with-cpp language mode, they can invoke the preprocessor with the options needed.
Applied as suggested:
> Matthias: I don't think we need to worry about other tools or the
assembly-with-cpp language mode; users can invoke the preprocessor with
the options needed.
> Likewise, this is missing the second half of the point I made and has
> a small error in it:
>> Matthias: Linux .S sources depend on $ not being a digit, but it can be worked around with additional macro expansion.
> I suggest
>> Matthias: Linux .S sources depend on $ not being a nondigit, but it can be worked around with additional macro expansion. Linux does however also depend on $ being treated as nondigit for some of their .c sources.
Thank you! I recall being confused about that error when editing after
the meeting and assumed I misunderstood something.
Applied as:
> Matthias: Linux |.S| sources depend on |$| not being a nondigit, but
it can be worked around with additional macro expansion; however, some
of the Linux |.c| sources do depend on |$| being treated as nondigit.
> In the following quote
>> Aaron: I think there were some recently reported issues involving $ in the preprocessor. [...]
>> Matthias: One of those was reported by me.
> Only one of the links is to an issue (which was not reported by me).
> The other one is a PR by me, which is also linked in the paper - I
> believe I did point out that it's just implementation experience for
> Option 3 of the paper.
Applied as:
> Matthias: One of those is a PR submitted by me; it's linked in the
paper and is implementation experience for the proposed option 3.
Tom.
>
>
> Best regards,
> Matthias
>
> On Mon, Jul 13, 2026 at 5:37 AM Tom Honermann<tom_at_[hidden]> wrote:
>> The consistent behavior across compilers that use Itanium mangling inspired me to submit a PR to the Itanium ABI to add $ to the <identifier> non-terminal.
>>
>> https://github.com/itanium-cxx-abi/cxx-abi/pull/213: Allow '$' in identifiers
>>
>> Tom.
>>
>> On 7/12/26 8:06 PM, Tom Honermann via SG16 wrote:
>>
>> Minutes have been published for the 2026-07-08 SG16 meeting here. For those that attended, please review and offer corrections.
>>
>> Tom.
>>
>>
Received on 2026-07-16 04:07:42