Went to Chester yesterday and drank some beer with some friends I hadn't seen for 20 years and now things are much better today!
I left it running pacman -Syuu yesterday and it fell over at QScintilla being corrupt as before. Ran it again today and it updated a shed load more packages, include QT again and didn't fall over at QScintilla.
make clean works again and make -j4 is four time faster, thanks for that.
Still have the snprintf problem. I tried reordering the includes but after changing loads of files I got stuck at OpenCSGWarningDialog.h because no order works.
snprintf doesn't seem to be used in OpenSCAD, so I undefed it after the include of linbintl.h and that solves the compile problem.
I still got the segfault at startup. I don't think Windows or bash add to the command line and even if they did it shouldn't segfault.
It did crash in our main() function in openscad.cc. It gets renamed by a #define to qMain. That is why I thought it was part of QT. Using #define to change function names should be banned!
Turns out I had some boost dlls in my openscad directory. Deleting those made it run again. Not sure why they were there from 2017 but they obviously need to match the headers in boost. That is why updating the packages made it crash.
So I am back in business with MSYS, so I can start to debug the Auto Reload problem.
> On Feb 18, 2019, at 06:15, nop head <[hidden email]> wrote:
> Since the crash is in boost called by Qt I don't see how it has anything to do with openscad.
Qt doesn't use boost, so this is most likely cmd-line handling in openscad.cc
Perhaps Windows passes some extra cmd-line argument to the process, triggering a failure to validate arguments on our side?
That exact issue has happened on Mac before..