Update to lrslib v7.1a

[blegat]

No description provided.

[blegat]

@oyamad Do you know what we should do in lrslib v7.1a for this ? With this line, it complains that FirstTime is not defined and without this line, I have a test failing (see the test logs, not sure if that's related to this line or if that's another change).

[oyamad]

Yes, I believe we still need it: there is no change in v7.1a: https://github.com/JuliaPolyhedra/lrslib/blob/ab1a5313a54a6d9aadcaa2f2e8523856f747f7e1/lrsnashlib.c#L55

Isn't this as pointed out by @bzinberg in JuliaPolyhedra/lrslib#2 (comment)?
The declaration of FirstTime has been removed in lrsnashlib.h in v7.1a: JuliaPolyhedra/lrslib@607673e#diff-8325ef3172d6193ffabfcb846d744f86c389a21a61c802c4c715f55462857204L70

See bzinberg/lrslib@453d0b0

[bzinberg]

Agreed. In case helpful, my thoughts as a newcomer to this codebase are:

• Changing the public API from an extern long that the user unsafe-sets to a reset() method is a good start. But probably more important, the re-implementation of lrs_solve_nash in Julia here seems like an anti-pattern to me. It indicates that FirstTime really isn't a public API, it's just needed to enable the implementation to be split across the C codebase and the Julia codebase. This is doable if it's needed, but perhaps we can revisit whether there's a way to bring all the logic into C with a documented API?
• FirstTime appears to be global state for the Nash solver. I wonder if that state could be factored into a struct so that the logic in lrsnashlib could become stateless?
• (I'm not sure why we're using a long for what apparently has the semantics of a boolean, but I assume there is a good reason for that.)

[blegat]

I agree. If the API was sufficient, we shouldn't have to define the structs in Julia. However, it's sometimes difficult to bring the API to that level to the upstream package.
We already contacted David Avis trying to merge the changes we need upstream but we probably complicated things asking to have the build process work on more platforms which is now resolved and simplified thanks to the cross-compilation with Yggdrasil. We could aim at getting
JuliaPolyhedra/lrslib@3962f83
JuliaPolyhedra/lrslib@ab1a531
and
JuliaPolyhedra/lrslib#3
to the next lrslib release and we could get rid of the fork.

[codecov]

Codecov Report

Merging #41 (364bb3a) into master (dba1279) will increase coverage by 0.15%.
The diff coverage is 91.66%.

@@            Coverage Diff             @@
##           master      #41      +/-   ##
==========================================
+ Coverage   83.70%   83.85%   +0.15%     
==========================================
  Files           8        8              
  Lines         632      638       +6     
==========================================
+ Hits          529      535       +6     
  Misses        103      103              

Impacted Files	Coverage Δ
src/LRSLib.jl	71.42% <0.00%> (ø)
src/lrstypes.jl	100.00% <ø> (ø)
src/matrix.jl	89.51% <100.00%> (+0.33%)	⬆️
src/nash.jl	92.30% <100.00%> (-0.06%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update dba1279...364bb3a. Read the comment docs.