https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92613
anlauf at gcc dot gnu.org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |anlauf at gcc dot gnu.org
--- Comment #2 from anlauf at gcc dot gnu.org ---
(In reply to Christopher Albert from comment #1)
> Created attachment 62992 [details]
> Fix bogus warning with -cpp -fpreprocessed
>
> This fixes the issue by skipping libcpp initialization and tokenization for
> normal compilations when both '-cpp' and '-fpreprocessed' are specified for
> Fortran code. Let me know if it works and what is your opinion on that.
A few comments:
- gfortran does preprocessing using cpp in the *traditional* mode, unlike
several other brands that are Fortran-aware.
- Fortran comments can start with "!", but in fixed-form also with the
corresponding comment character "c".
- The issue occurs with "'" as well as with '"'.
- How do you intend to have the testcase run? With -cpp -E, or without -E?
And what do you intend to test?
(Compiling with -E lets it fail as is, and without -E it is silent.
So what is the point of having the dg-bogus lines?)
I just didn't get the point.
Other comments:
- I find the comments in the source file a little excessive. Can you
formulate more concisely?
+When compiling a file that has already been fully preprocessed by an
"fully" is not needed (what does it mean?)
+! This comment has a "quote" too { dg-bogus "missing terminating" }
That is really bogus. Why should one ever get a warning at all?
Anyway, the patch is a good point to start the discussion with.
And I wonder if we should instead give a warning if someone combines
-cpp -fpreprocessed as contradicting options.
(Yes, I have seen the cmake issue).
