Bug #1661: Microsoft: cannot compile with signal handling - CoCoALib - CoCoA Project

Bug #1661

Microsoft: cannot compile with signal handling

Added by John Abbott about 2 years ago. Updated 17 days ago.

Status:

Closed

Priority:

Normal

Assignee:

Nico Mexis

Category:

Portability

Target version:

CoCoALib-0.99850

Start date:

09 Feb 2022

Due date:

% Done:

100%

Estimated time:

Spent time:

1.80 h

Description

Bruns reported by email a problem compiling on Microsoft:

my failed attempt to compile CoCoALib under MSYS MinGW64 for a Windows binary of Normaliz.
I go´t stuck at the signal handling. Can you suggest something?

Can we fix this?

Related issues

History

#1 Updated by John Abbott 2 months ago

Related to Design #1630: Signal handler not portable? added

#2 Updated by John Abbott 2 months ago

I have added Nico Mexis as a watcher. He has successfully built CoCoA-5 for Microsoft, so may have some useful insights!

#3 Updated by Nico Mexis 2 months ago

Status changed from New to In Progress
% Done changed from 0 to 50

Ah, thanks for making me aware of this issue. Actually, recently I put my instructions online for exactly that.
I had also tried to build CoCoA both through MSYS 2 and MinGW, but they both failed.
It works flawlessly with CygWin, though.
What's needed to fix other installations? Don't know exactly. Most likely it is because of the near-perfect POSIX layer of CygWin that it works so great.

Anyway, TL;DR, here is my gist: https://gist.github.com/ThexXTURBOXx/b75cd2e4721904465718b6c6298333e8

#4 Updated by John Abbott 2 months ago

Winfried sent me the following message by email:

I have compiled CoCoALib for MSYS MinGW64 about two years ago. It is used in the MS-Windows version of Normaliz
I also succeeded in making the other auxiliary packages available for MSYS MinGW 64. It was quite a struggle,
so I don't touch what I have.

If wanted I can provide more information.

I have asked him to send me the details.

#5 Updated by John Abbott about 1 month ago

% Done changed from 50 to 60

I have been working through Nico's changes. Still undecided about several of them :-/
The 32-bit vs 64-bit problems could perhaps be circumvented by using long long int or unsigned long long int in some places; but I fear that may become "cancerous". Nico put in some #ifdef macros to hide code which does not compile on 32-bit platforms. Maybe config.H needs to define such a macro: perhaps CoCoA_32bit and/or CoCoA_64bit

Time for a break!

#6 Updated by Nico Mexis about 1 month ago

A macro like CoCoA_32bit and/or CoCoA_64bit sounds good.
I also thought about converting some of the problematic values to [unsigned] long long int, but I was unsure of possible side effects. I also think I encountered some because the values were not convertible anymore using the functions from convert.H... But maybe I misremember that - I have tried too many things with the UCRT compiler...