[CRAN pretest] against v2.2.0, pgsc newly broken

[astamm]

Package: pgsc
Check: re-building of vignette outputs
New result: ERROR
 Error(s) in re-building vignettes:
   ...
 --- re-building ‘pgsc_vignette.Rmd’ using rmarkdown
 ! TeX capacity exceeded, sorry [main memory size=5000000].
 <argument> ...000, 0.191751, 0.203537, 0.166274, 0
                                                   .000000, 0.211288)
 l.24117 ....203537, 0.166274, 0.000000, 0.211288)}

 Error: processing vignette 'pgsc_vignette.Rmd' failed with diagnostics:
 LaTeX failed to compile /home/hornik/tmp/CRAN_recheck/pgsc.Rcheck/vign_test/pgsc/vignettes/pgsc_vignette.tex. See https://yihui.org/tinytex/r/#debugging for debugging tips. See pgsc_vignette.log for more info.
 --- failed re-building ‘pgsc_vignette.Rmd’

 SUMMARY: processing the following file failed:
   ‘pgsc_vignette.Rmd’

 Error: Vignette re-building failed.
 Execution halted

[astamm]

I ran R CMD check against the repo at cran/pgsc with the proposed nloptr v2.2.0 and ran into the same error

--- re-building ‘pgsc_vignette.Rmd’ using rmarkdown
   ! TeX capacity exceeded, sorry [main memory size=5000000].
   \FV@ListProcessLine ...\FV@RightListNumber }\hss }
   
   Error: processing vignette 'pgsc_vignette.Rmd' failed with diagnostics:
   LaTeX failed to compile /private/var/folders/sr/w672jz915x903zgl5485_n700000gn/T/Rtmp6zWutH/Rbuild15f6c604b6241/pgsc/vignettes/pgsc_vignette.tex. See https://yihui.org/tinytex/r/#debugging for debugging tips. See pgsc_vignette.log for more info.
   --- failed re-building ‘pgsc_vignette.Rmd’
   
   SUMMARY: processing the following file failed:
     ‘pgsc_vignette.Rmd’

So I guess it was a false positive too.

[astamm]

However running R CMD check against current CRAN version of nloptr goes well. So this is newly broken.

[astamm]

I found the issue which is that now print_level in nloptr.c is cast as unsigned int and not int.
In pgsc, they have a print.level argument in pgcs() function which is passed as option to the nloptr() function.
That argument in their vignette sometimes has the value -1 or -2 which is converted to +1 and +2 when cast to unsigned int, resulting in much more verbose that TeX can handle.

One easy option is to switch back to casting print_level as int in nloptr.c.
But on the other hand, print_level has no meaning (at least in nloptr) with negative values.

The other option would be to tell maintainer of pgsc to not use negative value but rather 0 if (s)he wants to turn verbose off. But they might use also print.level to output other messages non specific to nloptr.

@aadler: do you have a strong opinion about this?

[aadler]

@astamm No strong opinion at all. Thank you!

[astamm]

Then if he does not get back at me today, I'll proceed with my fix to get nloptr to CRAN.