[Bug c++/48839] New: #error should terminate compilation - similar to missing #include

[Bug c++/48839] New: #error should terminate compilation - similar to missing #include

[john.salmon at deshaw dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48839

           Summary: #error should terminate compilation - similar to
                    missing #include
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: john.salmon@deshaw.com


bug 15638 established the policy that a missing #include terminates a
compilation.

I would suggest that encountering an #error should be immediately fatal, in
just the same way, and for exactly the same reasons.

A specific use-case:  I've used something autoconf-like to tell me whether
C++0x-style auto is supported.  Then I write this in my source file:

#ifndef HAVE_CXX0X_AUTO
#error This file requires the auto feature of C++0x
#endif
// lots of code that uses auto.

When I compile it with -std=c++98, I get the initial error message, and then
thousands of lines of complaints about syntax errors in the rest of the file. 
It would be far better to stop after the #error.

[Bug preprocessor/48839] #error should terminate compilation - similar to missing #include

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48839

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at gcc dot gnu.org,
                   |                            |paolo.carlini at oracle dot
                   |                            |com
          Component|c++                         |preprocessor

--- Comment #1 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-09-25 11:07:49 UTC ---
Personally, I agree. Actually, I had to work around this behavior in the
library headers by changing the macros to something like:

#ifndef HAVE_CXX0X_AUTO
#error This file requires the auto feature of C++0x
#else
// lots of code that uses auto.
#endif

Note, this doesn't seem C++ specific, is a preprocessor issue. I'd like to hear
Jason about this.

[Bug preprocessor/48839] #error should terminate compilation - similar to missing #include

[egall at gwmail dot gwu.edu]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=48839

Eric Gallager <egall at gwmail dot gwu.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |egall at gwmail dot gwu.edu

--- Comment #2 from Eric Gallager <egall at gwmail dot gwu.edu> ---
I have been working around this by using the -Wfatal-errors flag when relevant,
but I agree that an actual fix would be better than the workaround...

[Bug preprocessor/48839] #error should terminate compilation - similar to missing #include

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=48839

--- Comment #3 from Jason Merrill <jason at gcc dot gnu.org> ---
Makes sense to me.

[Bug preprocessor/48839] #error should terminate compilation - similar to missing #include

[paolo.carlini at oracle dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=48839

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |joseph at codesourcery dot com

--- Comment #4 from Paolo Carlini <paolo.carlini at oracle dot com> ---
Thanks. I think the patch proper could be something as simple as the below, but
quite a few C testcases would require updating. I'd like to hear Joseph too
before the next steps.

Index: directives.c
===================================================================
--- directives.c    (revision 219024)
+++ directives.c    (working copy)
@@ -1103,7 +1103,7 @@ do_diagnostic (cpp_reader *pfile, int code, int re
 static void
 do_error (cpp_reader *pfile)
 {
-  do_diagnostic (pfile, CPP_DL_ERROR, 0, 1);
+  do_diagnostic (pfile, CPP_DL_FATAL, 0, 1);
 }

 static void

[Bug preprocessor/48839] #error should terminate compilation - similar to missing #include

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=48839

--- Comment #11 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Paolo Carlini from comment #8)
> about the other compilers? For sure some are following the current gcc
> behaviour, for compatibility, I suspect ICC for example. Should we also have
> a command line switch? Probably missing #includes just unconditionally end
> the compilation on most compilers, though. Well, it's also true that the so
> called workaround which I mentioned in Comment #1 works pretty well and we
> do have *very* few fatal errors in the front ends...

Clang keeps going, like GCC. ICC seems to as well.

But EDG stops immediately with a "catastrophic error":

"foo.cc", line 1: catastrophic error: #error directive: NOPE


Anyway, I don't think compatibility is an issue. We're not going to produce any
assembly or other output, compilation fails.

[Bug preprocessor/48839] #error should terminate compilation - similar to missing #include

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=48839

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|paolo.carlini at oracle dot com    |unassigned at gcc dot gnu.org
             Status|ASSIGNED                    |NEW

--- Comment #12 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Paolo has not been around for the last 3 years so unassigning this bug report
from him so someone else can take it if they want. The patch against libcpp is
simple but the testsuite needs to updated or improved.
It also might make sense to have this as an option instead; though
-Wfatal-errors might be enough for that ... (maybe just an improved
documentation for #error in the cpp manual is needed for this change in the
end).

[Bug preprocessor/48839] #error should terminate compilation - similar to missing #include

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=48839

--- Comment #13 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note C23 has the following under "4. Conformance":
The implementation shall not successfully translate a preprocessing translation
unit containing a
#error preprocessing directive unless it is part of a group skipped by
conditional inclusion.

GCC does error out when preprocessing seperately from compiling but when doing
a combined, GCC does "successfully" translate the preprocessing translation but
does not finish compiling though. But maybe that is still valid.