[Groff] Re: pre-grohtml 1.18/Win32

[Werner LEMBERG]

[About porting groff to the MKS environment on Win32]

> Attached are the revised pre-html as well as the output of diff -c
> (I don't have 'u') vs. today's snapshot of pre-html.cc.  Attached
> also is a revised configure script and a revised Makefile.in (some
> of the tmacpath definitions are peculiar to my system).

This looks very promising!  Do you have time to update your changes
to the current version of pre-html.cc?

>     MKS Toolkit for Developers 8.1, including the MKS port of GNU
>     make (I'm not sure which version)

Say `make --version'.

> The revised configure script was invoked as follows:
> 
> configure -C CC=cl CFLAGS='-O2 -G6 -nologo' \
>              CXX=cl CXXFLAGS='-O2 -G6 -nologo' \
>              CXXEXT=cpp LIBEXT=lib ac_exeext=.exe

At least specifying LIBEXT shouldn't be necessary as soon as MKS is
recognized by config.guess; I will then add support to groff.

Instead of using CXXEXT I now tend to rename all C++ files to *.cpp --
it seems that virtually all modern C++ compilers support that
extension out of the box (including the *very* strange compiler on
z/OS).  Any objections?

> FIX: Add to config.hin:
> /* Define if you have the `isatty' function. */
> #undef HAVE_ISATTY

Done.

> tmpfile.cc(137) : error C2664: 'delete' : cannot convert parameter 1
>   from 'const char *' to 'void *' Conversion loses qualifiers
> 
> FIX: Add cast:
> a_delete (void *) tmp->fname;
> Note: I don't really know C++, so this was a bit of guess.

I tried a different solution (using `char*' instead of `const char*'
for xtmpfile_list.fname).  Please test.

> tmpname.cc(84) : error C2065: 'time' : undeclared identifier
> FIX: Add:
> #include <time.h>

Done.

> input.cc(847) : error C2666: '==' : 5 overloads have similar conversions
> ...
> 
> FIX: Add (int) casts
> Note: as with cast in tmpfile.cc, this was a bit of guess.

I've added proper operator definitions for `==' and `!=' in the `Char'
class to compare the `char' type with `Char', making the casts
unnecessary (I think).  Again, please test.

> Of course, without yacc, tbl, eqn, pic, and refer don't compile
> either.

Haven't you received the archive with the processed yacc files?


    Werner