[Bug preprocessor/95889] New: corrupt size vs. prev_size and hang when using __has_include with -traditional-cpp

[Bug preprocessor/95889] New: corrupt size vs. prev_size and hang when using __has_include with -traditional-cpp

[dev-zero at gentoo dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95889

            Bug ID: 95889
           Summary: corrupt size vs. prev_size and hang when using
                    __has_include with -traditional-cpp
           Product: gcc
           Version: 10.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: preprocessor
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dev-zero at gentoo dot org
  Target Milestone: ---

Part of my testing for issues like https://github.com/evaleev/libint/issues/173
with GCC 10.1 I came up with a silly test header:

****snip****
#if defined __has_include
#  if __has_include(<stdio.h>)
hello there
#  endif
#endif
****snip****

which works as expected with `cpp test.h`:

****snip****
# 1 "test.h"
# 1 "<built-in>"
# 1 "<command-line>"
# 31 "<command-line>"
# 1 "/usr/include/stdc-predef.h" 1 3 4
# 32 "<command-line>" 2
# 1 "test.h"



hello there
****snip****

But gives the following and hangs when running with `cpp -traditional-cpp
test.h`:

****snip****
# 1 "test.h"
# 1 "<built-in>"
# 1 "<command-line>"
# 31 "<command-line>"
# 1 "/usr/include/stdc-predef.h" 1 3 4

# 17 "/usr/include/stdc-predef.h" 3 4

[...]

# 32 "<command-line>" 2
# 1 "test.h"
test.h:3: error: missing '(' before "__has_include" operand
    3 | #  if __has_include(<stdio.h>)
      | 
test.h:3: error: operator "__has_include" requires a header-name
test.h:3: warning: null character(s) ignored
test.h:3: error: missing binary operator before token "q"
*** Error in
`/data/tiziano/cp2k/tools/toolchain/install/gcc-10.1.0/lib/gcc/x86_64-pc-linux-gnu/10.1.0/cc1':
corrupted size vs. prev_size: 0x0000000002a517a0 ***
test.h:1: confused by earlier errors, bailing out
****snip****

Besides this issue with the compiler itself, I wonder what the correct
invocation of `__has_include` in traditional mode is (because it seems to be
defined).

[Bug preprocessor/95889] corrupt size vs. prev_size and hang when using __has_include with -traditional-cpp

[dev-zero at gentoo dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95889

--- Comment #1 from Tiziano Müller <dev-zero at gentoo dot org> ---
Unfortunately `__has_include` is even being parsed (in traditional-cpp mode) if
enclosed in a `if 0 ... endif`:

****snip****
#if 0
#if __has_include(<foo.h>)
#endif
#endif
****snip****

which gives:

****snip****
# 32 "<command-line>" 2
# 1 "test2.h"
test2.h:3: error: missing '(' before "__has_include" operand
    3 | #if __has_include(<foo.h>)
      | 
test2.h:3: error: operator "__has_include" requires a header-name
*** Error in
`/data/tiziano/cp2k/tools/toolchain/install/gcc-10.1.0/lib/gcc/x86_64-pc-linux-gnu/10.1.0/cc1':
corrupted size vs. prev_size: 0x000000000249a7e0 ***
test2.h:1: confused by earlier errors, bailing out
****snip****

while enclosing it in parenthesis at least fixes that, such that the following
parses fine:

****snip****
#if 0
#if (__has_include(<foo.h>))
#endif
#endif
****snip****

[Bug preprocessor/95889] __has_include evaluated even if disabled (with cpp in traditional-mode)

[dev-zero at gentoo dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95889

Tiziano Müller <dev-zero at gentoo dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|10.1.0                      |10.2.0
            Summary|corrupt size vs. prev_size  |__has_include evaluated
                   |and hang when using         |even if disabled (with cpp
                   |__has_include with          |in traditional-mode)
                   |-traditional-cpp            |
      Known to work|                            |9.3.0

--- Comment #2 from Tiziano Müller <dev-zero at gentoo dot org> ---
Since the ICE is gone in gcc-10.2, there is only the second issue left:
__has_include is evaluated even though it occurs in a "disabled region" when
running cpp in traditional mode. This breaks building some Fortran packages,
even when they guard the __has_include stmt with a `#if defined __has_include`
as per the documentation.

test2.h:

  #if 0
  #if (__has_include(<foo.h>))
  #endif
  #endif

$ ./install/gcc-10.2.0/bin/cpp -traditional-cpp test2.h
# 1 "test2.h"
# 1 "<built-in>"
# 1 "<command-line>"
# 31 "<command-line>"
# 1 "/usr/include/stdc-predef.h" 1 3 4
# 17 "/usr/include/stdc-predef.h" 3 4
# 32 "<command-line>" 2
# 1 "test2.h"
test2.h:3: error: operator "__has_include" requires a header-name
    3 | #if (__has_include(<foo.h>))
      |

[Bug preprocessor/95889] __has_include evaluated even if disabled (with cpp in traditional-mode)

[tm@dev-zero.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95889

--- Comment #3 from Tiziano Müller <tm@dev-zero.ch> ---
Created attachment 48941
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=48941&action=edit
fix __has_include with traditional-cpp

as posted on the gcc-patches ml

[Bug preprocessor/95889] [10 regression] __has_include broken with -traditional-cpp (and therefore with Fortran)

[tm@dev-zero.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95889

Tiziano Müller <tm@dev-zero.ch> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|__has_include evaluated     |[10 regression]
                   |even if disabled (with cpp  |__has_include broken with
                   |in traditional-mode)        |-traditional-cpp (and
                   |                            |therefore with Fortran)
                 CC|                            |jakub at gcc dot gnu.org,
                   |                            |nathan at acm dot org

--- Comment #4 from Tiziano Müller <tm@dev-zero.ch> ---
Link to patch series here:
https://gcc.gnu.org/pipermail/gcc-patches/2020-July/550832.html

[Bug preprocessor/95889] [10 regression] __has_include broken with -traditional-cpp (and therefore with Fortran)

[nathan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95889

Nathan Sidwell <nathan at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |nathan at gcc dot gnu.org
   Last reconfirmed|                            |2020-07-29
             Status|UNCONFIRMED                 |ASSIGNED
                 CC|                            |nathan at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #5 from Nathan Sidwell <nathan at gcc dot gnu.org> ---
thanks for the patch, taking so I don't forget

[Bug preprocessor/95889] [10 regression] __has_include broken with -traditional-cpp (and therefore with Fortran)

[nathan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95889

Nathan Sidwell <nathan at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #6 from Nathan Sidwell <nathan at gcc dot gnu.org> ---
Fixed trunk f6fe3bbf9f6
Fixed gcc-10 5e66b6c0103