[Bug c++/107351] New: Bootstrap fails when gcc is built with gcc-13-master: error: conversion from 'long unsigned int' to 'unsigned int' changes value from '7001010000' to '2706042704'

[Bug c++/107351] New: Bootstrap fails when gcc is built with gcc-13-master: error: conversion from 'long unsigned int' to 'unsigned int' changes value from '7001010000' to '2706042704'

[slyfox at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107351

            Bug ID: 107351
           Summary: Bootstrap fails when gcc is built with gcc-13-master:
                    error: conversion from 'long unsigned int' to
                    'unsigned int' changes value from '7001010000' to
                    '2706042704'
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: slyfox at gcc dot gnu.org
  Target Milestone: ---

When building gcc-13-master with gcc-13-master I see the following bootstrap
failure:

../../source/gcc/cp/module.cc: In member function 'void
module_state::write_readme(elf_out*, cpp_reader*, const char*)':
<command-line>: error: conversion from 'long unsigned int' to 'unsigned int'
changes value from '7001010000' to '2706042704' [-Werror=overflow]
../../source/gcc/cp/module.cc:14170:19: note: in expansion of macro
'MODULE_VERSION'
14170 |   version2string (MODULE_VERSION, string);
      |                   ^~~~~~~~~~~~~~

I understand bootstrap with unreleased gcc is not very well supported. But in
this case it sounds like a natural fix. Worth switching to `uint64_t` to store
gcc version internally?

configure flags: --prefix=/<<NIX>>/gcc-13.0.0
--with-gmp-include=/<<NIX>>/gmp-with-cxx-6.2.1-dev/include
--with-gmp-lib=/<<NIX>>/gmp-with-cxx-6.2.1/lib
--with-mpfr-include=/<<NIX>>/mpfr-4.1.0-dev/include
--with-mpfr-lib=/<<NIX>>/mpfr-4.1.0/lib --with-mpc=/<<NIX>>/libmpc-1.2.1
--with-native-system-header-dir=/<<NIX>>/glibc-2.35-163-dev/include
--with-build-sysroot=/ --program-prefix= --enable-lto --disable-libstdcxx-pch
--without-included-gettext --with-system-zlib --enable-checking=release
--enable-static --enable-languages=c\,c++ --disable-multilib --enable-plugin
--with-isl=/<<NIX>>/isl-0.20 --build=x86_64-unknown-linux-gnu
--host=x86_64-unknown-linux-gnu --target=x86_64-unknown-linux-gnu

[Bug c++/107351] Bootstrap fails when gcc is built with gcc-13-master: error: conversion from 'long unsigned int' to 'unsigned int' changes value from '7001010000' to '2706042704'

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107351

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2022-10-21
             Status|UNCONFIRMED                 |WAITING
     Ever confirmed|0                           |1

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
# In non-release builds, use a date-related module version.
ifneq ($(DEVPHASE_c),)
# Some date's don't grok 'r', if so, simply use today's
# date (don't bootstrap at midnight).
MODULE_VERSION := $(shell date -r $(srcdir)/cp/module.cc '+%y%m%d-%H%M' \
  2>/dev/null || date '+%y%m%d-0000' 2>/dev/null || echo 0)

CFLAGS-cp/module.o += -DMODULE_VERSION='($(subst -,,$(MODULE_VERSION))U)'
endif


I am trying to figure out how you would get 7001010000 if the above was being
used.
Can you attach the full build log around the building of module.cc ?

it should have been -DMODULE_VERSION=2210211757U or something similar.

Even in the source we have:
#define IS_EXPERIMENTAL(V) ((V) >= (1U << 20))
#define MODULE_MAJOR(V) ((V) / 10000)
#define MODULE_MINOR(V) ((V) % 10000)
#define EXPERIMENT(A,B) (IS_EXPERIMENTAL (MODULE_VERSION) ? (A) : (B))
#ifndef MODULE_VERSION
#include "bversion.h"
#define MODULE_VERSION (BUILDING_GCC_MAJOR * 10000U + BUILDING_GCC_MINOR)
#elif !IS_EXPERIMENTAL (MODULE_VERSION)
#error "This is not the version I was looking for."
#endif

[Bug c++/107351] Bootstrap fails when gcc is built with gcc-13-master: error: conversion from 'long unsigned int' to 'unsigned int' changes value from '7001010000' to '2706042704'

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107351

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Perhaps if the filesystem doesn't have timestamps or the file has for whatever
reason the Unix era timestamp (1970-01-01 00:00)?
The version fits into unsigned int only if it is in between 2000 and 2042
inclusive.

[Bug c++/107351] Bootstrap fails when gcc is built with gcc-13-master: error: conversion from 'long unsigned int' to 'unsigned int' changes value from '7001010000' to '2706042704'

[slyfox at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107351

--- Comment #3 from Sergei Trofimovich <slyfox at gcc dot gnu.org> ---
Created attachment 53750
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53750&action=edit
gcc-13.0.0-build.log.xz

[Bug c++/107351] Bootstrap fails when gcc is built with gcc-13-master: error: conversion from 'long unsigned int' to 'unsigned int' changes value from '7001010000' to '2706042704'

[slyfox at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107351

--- Comment #4 from Sergei Trofimovich <slyfox at gcc dot gnu.org> ---
Attached the build log.

I completely forgot about the fact that I was using gcc source from
"normalized" filesystem which intentionally resets all timestamps (it's one of
the build modes for my distribution to validate build determinism):

$ LANG=C ls -l
/nix/store/p2m8xz3klafr1ydbjf8g86vmylziscr3-source/gcc/cp/module.cc
-r--r--r-- 4 root root 533735 Jan  1  1970
/nix/store/p2m8xz3klafr1ydbjf8g86vmylziscr3-source/gcc/cp/module.cc

Silly question: would it be reasonable for module.cc to use
gcc/{BASE-VER,DATESTAMP}?

[Bug c++/107351] Bootstrap fails when gcc is built with gcc-13-master: error: conversion from 'long unsigned int' to 'unsigned int' changes value from '7001010000' to '2706042704'

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107351

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
-DMODULE_VERSION='(7001010000U)'

Hmm.
The date command is not listed in the build log either (debugging make here is
going to be "hard").

[Bug c++/107351] Bootstrap fails when gcc is built with gcc-13-master: error: conversion from 'long unsigned int' to 'unsigned int' changes value from '7001010000' to '2706042704'

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107351

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Sergei Trofimovich from comment #4)
> Attached the build log.
> 
> I completely forgot about the fact that I was using gcc source from
> "normalized" filesystem which intentionally resets all timestamps (it's one
> of the build modes for my distribution to validate build determinism):
> 
> $ LANG=C ls -l
> /nix/store/p2m8xz3klafr1ydbjf8g86vmylziscr3-source/gcc/cp/module.cc
> -r--r--r-- 4 root root 533735 Jan  1  1970
> /nix/store/p2m8xz3klafr1ydbjf8g86vmylziscr3-source/gcc/cp/module.cc
> 
> Silly question: would it be reasonable for module.cc to use
> gcc/{BASE-VER,DATESTAMP}?

The problem with DATESTAMP is that it changes every day while module support
does not. So this might be ok.
You might be one of the few which does the reset timestamp stuff and even then
it seems like it could be broken ...