On Sun, 14 May 2023, Sven Schreiber wrote:
Am 14.05.2023 um 16:07 schrieb Cottrell, Allin:
> I can confirm that, and I'm seeing this:
>
> ' matrices c...'
> *** buffer overflow detected ***: terminated
> Aborted (core dumped)
>
> It's due to the massively long statements defining arrays of
> critical-value matrices. A fix should be forthcoming shortly, but I
> don't think such statements are the right way to store big matrices.
> Better to put them into .bin matrix files and load them on demand.
Well, maybe. I admit it was a lazy solution (copied over from the
original Ox code), but apart from the fact that it has "just
worked", the advantage is that a user can inspect those tables
quite easily. If we store the tables as bin matrices, it seems to
me we would hide them away. Of course, in principle we could
reproduce them in the help doc. Or would there be another way of
making them accessible - I mean apart from telling people to
manually poke around the directory where the file is stored and
then using mread() ?
I wouldn't say the matrices were exactly user-legible as things
stand. If it's worthwhile to make them user-accessible one could add
a function to show and/or return them.
However, I'm not saying you have to put them into .bin files. I just
noted that's what Ignacio does with his GHegy package (although he
uses mat.gz format) and it seems to me a good way of handling
constant auxiliary matrices.
Allin