[3dprinter-general] Bug#1037614: cura-engine: ftbfs with GCC-13

[Gregor Riepl]

Hi Christoph,

> There's still an issue on i386 left, and at least from looking at the
> CI tests, it might have been caused by this fix:
> 
> https://salsa.debian.org/3dprinting-team/cura-engine/-/pipelines
> 
> Any ideas?

It's actually this issue again:
https://github.com/Ultimaker/CuraEngine/issues/1192

I investigated this a bit more thoroughly, and I think it's caused by 
implicit truncation of double to long long, that can result in 
off-by-one errors when the x87 FPU produces a subtly different result 
than other floating point processors. This can quickly lead to an 
accumulation of errors, when lots of off-by-one results are added together.

I have attached a preliminary patch for the failing IntPointTest that 
uses explicit round-to-nearest instead of truncation, but nothing for 
the other issue yet. A quick workaround would be a patch like the one 
you posted in that bug report, but I think this is really a fundamental 
issue in CuraEngine that needs to be fixed properly.

Regards,
Gregor
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0017-Round-from-double-to-cInt-explicitly-when-applying-m.patch
Type: text/x-patch
Size: 1414 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/3dprinter-general/attachments/20230719/b708ea14/attachment.bin>