RE: BUG #16920: Can't compile PostGIS with MingW64 against PostgreSQL 14 head

Regina Obe <lr@pcorp.us>

From: "Regina Obe" <lr@pcorp.us>
To: "'Tom Lane'" <tgl@sss.pgh.pa.us>, "'Heikki Linnakangas'" <hlinnaka@iki.fi>
Cc: 'Juan José Santamaría Flecha' <juanjo.santamaria@gmail.com>, "'PostgreSQL mailing lists'" <pgsql-bugs@lists.postgresql.org>
Date: 2021-03-15T00:31:06Z
Lists: pgsql-bugs
> I came across a very relevant mingw bug report:
> 
> https://sourceforge.net/p/mingw-w64/bugs/465/
> 
> in which they were trying to clear an 8-byte jmp_buf field with a 4-byte
> instruction.  That would go a long way towards explaining the sort of
> problems we're seeing ... except it was fixed five years ago.
> I trust none of the buildfarm animals are still using old mingw versions.
> 
> 			regards, tom lane

The mingw64 gcc 8.1.0/msys2 I am using granted is not a spring chicken.  I
don't think it's 5 years old, but could be.

That said I did download the latest mingw64/msys2 package:

gcc.exe (Rev6, Built by MSYS2 project) 10.2.0

and I encountered the same crashers using the same configure when compiling.

Maybe it has more to do with the MSVCRT which mingw just uses what is on the
system - C:\WINDOWS\System32\msvcrt.dll
I'm running windows 10 which should have a fairly recent MSVCRT.  I wonder
if things would be different on windows 2019.


Thanks,
Regina






Commits

  1. Work around issues in MinGW-64's setjmp/longjmp support.