public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug libfortran/101305] New: Bind(C): Problems with incorrect kinds/sizes in ISO_Fortran_binding.h and CFI_establish
@ 2021-07-03  5:48 sandra at gcc dot gnu.org
  2021-07-06 20:11 ` [Bug libfortran/101305] " sandra at gcc dot gnu.org
                   ` (8 more replies)
  0 siblings, 9 replies; 10+ messages in thread
From: sandra at gcc dot gnu.org @ 2021-07-03  5:48 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101305

            Bug ID: 101305
           Summary: Bind(C): Problems with incorrect kinds/sizes in
                    ISO_Fortran_binding.h and CFI_establish
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libfortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: sandra at gcc dot gnu.org
                CC: burnus at gcc dot gnu.org, jrfsousa at gcc dot gnu.org
  Target Milestone: ---

ISO_Fortran_binding.h incorrectly hard-codes literal data sizes into the
CFI_type_* macros that are only appropriate on a 64-bit target, such as
CFI_type_ptrdiff_t having the same kind (8) as CFI_type_int64_t.  Even on
x86_64-linux-gnu, CFI_type_int_fast16_t and CFI_type_int_fast32_t have kinds
that are inconsistent with the sizes of the corresponding types as defined in
the system headers.

In addition, the standard says:  "If a C type is not interoperable with a
Fortran type and kind supported by the Fortran processor, its macro shall
evaluate to a negative value."  ISO_Fortran_binding.h doesn't handle that (I
think this mainly affects the 128-bit types GCC supports as an extension on
some targets).

There are related problems in CFI_establish in filling in the elem_len in the
descriptor from the kind.  It incorrectly ends up giving zero elem_len to
CFI_type_cptr and CFI_type_cfunptr, and explicitly gives size 64(!) to long
double kind 10.  It should probably reject unsupported types with
CFI_INVALID_TYPE.

There are test cases in the WIP TS 29113 testsuite:

https://gcc.gnu.org/pipermail/gcc-patches/2021-July/574115.html

interoperability/typecodes-array.f90
interoperability/typecodes-array-ext.f90
interoperability/typecodes-scalar.f90
interoperability/typecodes-scalar-ext.f90
library/establish.f90

This issue has some overlap with the set of bugs addressed by this patch from
José (PR fortran/100906/100907/100911/100914/100915/100916):

https://gcc.gnu.org/pipermail/fortran/2021-June/056154.html

but that patch is focused on the Fortran side of things rather than the C side.

Tobias and I both have WIP patches for this issue that need to be merged (his
is for handling detection of Fortran processor support, mine is to use sizeof 
instead of hard-coding sizes).  Probably all 3 pieces need to be in place for
this to work correctly.

^ permalink raw reply	[flat|nested] 10+ messages in thread
end of thread, other threads:[~2021-08-10 15:27 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-07-03  5:48 [Bug libfortran/101305] New: Bind(C): Problems with incorrect kinds/sizes in ISO_Fortran_binding.h and CFI_establish sandra at gcc dot gnu.org
2021-07-06 20:11 ` [Bug libfortran/101305] " sandra at gcc dot gnu.org
2021-07-08 19:03 ` sandra at gcc dot gnu.org
2021-07-14  0:08 ` sandra at gcc dot gnu.org
2021-07-28  4:24 ` cvs-commit at gcc dot gnu.org
2021-07-28  4:24 ` cvs-commit at gcc dot gnu.org
2021-07-28  4:24 ` cvs-commit at gcc dot gnu.org
2021-07-28 20:21 ` sandra at gcc dot gnu.org
2021-08-09 10:36 ` cvs-commit at gcc dot gnu.org
2021-08-10 15:27 ` cvs-commit at gcc dot gnu.org

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).