[cfe-dev] Bug 39206 - [Modules TS] Macro defined in module interface cannot be set in implementation
Matt Asplund via cfe-dev
cfe-dev at lists.llvm.org
Sun Oct 21 19:08:22 PDT 2018
I believe there is an inherent issue with how clang uses the precompiled
header structure to implement a module interface definition. When linking
against this module file clang brings in all of the header lookup, which I
can fix by forcing ModulesTS to not include the external source. However,
it feels like I am fighting the pcm to ignore already included files and
previously defined symbols that should have global module linkage. From
reading the spec it feels like the pcm file should ONLY contain the symbols
available from name lookup from module interface purview to implementation
purview as well as exported symbols that are imported.
Side note: As I try to track down these issues it seems odd that ModulesTS
is forcing Modules to be turned on as well. Would it be easier to keep
Clang Modules entirely separate from ModulesTS?
-Matt
On Sun, Oct 21, 2018 at 5:54 PM Matt Asplund <mwasplund at gmail.com> wrote:
> This was discussed in a separate thread, but got lost in the discussion.
>
> I am currently working on porting a legacy project over to using the new
> modules-ts and would hit the issue linked in *Bug 39206
> <https://bugs.llvm.org/show_bug.cgi?id=39206> *I could not find any
> explicit section of the spec that says the processor should be effected by
> compilation of a module, however clang seems to be hiding macros that were
> defined in included header files that were also included in the interface
> module.
>
> If this is indeed a bug I would also appreciate help pointing in the right
> direction in the code. I have spent some time trying to find where the
> actual check is occurring that is hiding macros, but am at a loss.
>
> Thanks,
> Matt
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20181021/43759319/attachment.html>
More information about the cfe-dev
mailing list