[Boris]
> Sure, nothing prevents you from saving the transcripts of the module
> mapper exchange. In fact, build2 does this at a sufficiently high
> verbosity levels along with the command lines to allow the user to
> replay the mapper replies, say, for troubleshooting.

Thanks. Good to know. It would be for production use, not just troubleshooting.

> Sure, module mapper approach is not without its complications (I
> even wrote a paper on this). But so far the only proposed alternative
> is to pre-scan the world, which doesn't feel scalable (and, as you
> know, I have doubts it can be implemented correctly for header units).

Pre-scan the world is a bit of an exaggeration though. Once the library dependencies are specified, it is a pretty fast process.
MSBuild has deployed. I suspect CMake is taking similar approach, but I wouldn't want to put words in the mouth of CMake developers.

-- Gaby

-----Original Message-----
From: Boris Kolpackov <boris_at_[hidden]>
Sent: Friday, April 15, 2022 4:42 AM
To: Gabriel Dos Reis <gdr_at_[hidden]>
Cc: ext_at_[hidden]; Ben Boeckel <ben.boeckel_at_[hidden]>; WG21 Tooling Study Group SG15 <tooling_at_[hidden]>
Subject: Re: [isocpp-ext] Can we expect that all C++ source files can have the same suffix?

Gabriel Dos Reis <gdr_at_[hidden]> writes:

> There are environments where the static inputs to the builds and
> a transcript of the decision making process need to be persisted
> for post-build auditing (was the build subverted in unsuspected
> ways?); that is fundamental and non-negotiable in those setups.

Sure, nothing prevents you from saving the transcripts of the module
mapper exchange. In fact, build2 does this at a sufficiently high
verbosity levels along with the command lines to allow the user to
replay the mapper replies, say, for troubleshooting.


> I do find values in a module mapper where an interactive
> development (like active IDE development) can do with dynamic
> queries. But I am skeptical that it is the answer for all.

Sure, module mapper approach is not without its complications (I
even wrote a paper on this). But so far the only proposed alternative
is to pre-scan the world, which doesn't feel scalable (and, as you
know, I have doubts it can be implemented correctly for header units).