https://bugs.winehq.org/show_bug.cgi?id=47592
--- Comment #13 from Jean-Michel Friedt friedtj@free.fr ---
So that proves there's a bug somewhere in our native components.
I'd love to help correcting this issue. At least I can confirm that disabling the MS libraries and getting back to the Wine libraries produced the initial effect, namely a slight error on the antenna reflection coefficient and a completely erroneous calculation on the radiation pattern and radiation efficiency (which I believe to be post-processing steps).
Of course we'd like to be able to fix whatever's broken with our builtin libraries, since relying on Microsoft redistributables is not great for several reasons. If you have the time, then, could you please reset the libraries back to builtin (open winecfg and remove overrides from the "libraries" tab), and then take a log with WINEDEBUG=+relay,+msvcrt,+vcomp? It'll probably take a lot longer, and produce a large output file, if it doesn't get prohibitively large you can compress and attach it here, or upload it elsewhere if it's too large even then.
12 GB is as prohibitively large as I can consider a dump, and the calculation was only starting so I am not even sure the erroneous part of the computation was reached (I am not well aware of the details of HFSS but my understanding is that is starts meshing space to prepare the simulation ... the log file ended up filling the space left on my hard disk by that time -- meaning we did not even get started with actually solving the problem yet). I am putting the part of the dump file that I could save (compressed 12 GB) at https://filesender.renater.fr/?s=download&token=7cd7d4f6-fbe2-432d-b2b5-... although the compressed log is only 250 MB large. Ideally I could try to simplify the problem to help isolate the computational error, but I am not sure I understand HFSS well enough not to simplify the model beyond reasonable.
I don't mind running the dump further on an external hard disk with more space available if it can be of any use -- if you have time answering my requests, I of course have time generating dump files -- but I am not sure such huge files can be of any use. If they can, I'll be happy generating more of them ! Also if there is an iterative way of testing one library after another rather than all at the same time, maybe we could converge faster.
Thanks, Jean-Michel