Re: gnucap save prompt error

[Felix Salfelder]

On Thu, Aug 27, 2020 at 10:01:30AM +0200, patrick wrote:
> > a wild guess
> > ...
> > +  assert(!is_file());
> 
> Thanks Felix. Yes, this works. InterestingI forgot to check the
> declaration for the variable. I will keep this in mind.

Al has fixed it here [1]. will hopefully show up in develop soon.

> > NB: I noticed that cmake does not either build in production nor in debug 
> > mode.
> > This is always suboptimal. In production mode, you would not have noticed 
> > the
> > assertion (-DNDEBUG). and in debug mode, the "untested" would have appeared 
> > on
> > the screen (-DTRACE_UNTESTED).
> >
> 
> I am able to use gdb --args gnucap ... to see a stacktrace with some
> symbols and file names in my CMake build. Probably the debug info you
> mention is something else.

You can see the upstream/intended difference between debug and release
builds in `diff lib/Make2 lib/Make2.Debug`.

-DNDEBUG is set in release mode only. this makes the program run (much)
faster. same for -O2. in debug builds, we set -O0 instead and turn on
debugging symbols (-g) and testing (-DTRACE_UNTESTED). In debug builds,
more warnings are enabled (-W*).

> Regarding two release types, you mean to have a CMake variable set
> like this 
> https://stackoverflow.com/questions/7724569/debug-vs-release-in-cmake

in upstream, you type "make debug" after configure, in autotools, there
is an --enable-debug switch in configure. I don't know what the best
practice in cmake is, but it looks like your link to so.com is close.

> By the way, what do you think would be important when loading
> different MOS models in Gnucap, e.g. from here [..]
> http://bsim.berkeley.edu/models/

In the absence of a reasonable verilog/A model compiler, using those
models can be difficult. I managed to run some of the BSIM family models
through gnucap-adms... gnucap-modelgen-verilog is meant to solve this,
but that is a longer story.

cheers
felix