Plan9 port to Bill-Gray fork

[staalmannen]

Dear, I saw that you had far more target platforms than upstream PDCurses so I was thinking that you might be interested in Plan9 too...

[GitMensch]

I like that idea in general. In any case this likely still needs some time to be in a state where I'd find it possible to pull (which is of course @Bill-Gray's decision in any case).

I'll drop some notes in the suggested file changes in the meantime.

[GitMensch]

The general approach looks fine. I suggest to first check if there's a possible CI build option for Plan9/pcc, otherwise this port may be broken now or at any time later.

Some documentation stuff (starting with comments, extending to the Makefile and C headers, ending in the README) should be updated before the port, too.

Question @staalmannen: did you recheck all the demos? Can you share screenshots?
Do you know if there was a PR to PDCurses before and if yes can reference it?

[GitMensch]

misses the newer demos, newtest, speed and possibly picsview

[staalmannen]

speed could not be built since I am missing sys/timeb.h

[GitMensch]

Can you check with an option to build without sys/timeb.h? Maybe @benavento has an idea?

[Bill-Gray]

sys/timeb.h is required by ftime(), for which the man page says

This function is obsolete.  Don't use it... gettimeofday(2)
gives microseconds;...

(Which is my fault; I wrote speed.c.) I'll revise the code to use gettimeofday().

[Bill-Gray]

Should work now (see commit 0b34997). Turns out my use of ftime() was not totally stupid; in fact, that was the only solution that works on some older compilers, and we can't get rid of it completely. But it's now only used on those older compilers, and it should compile/run correctly everywhere, including Plan9.

[GitMensch]

@staalmannen can you please update from master, add the speed demo and test it on plan9?

[staalmannen]

speed now builds (I got 19 fps). I had to disable wchar.h since that is not present by default, but it seems like nothing in speed.c depended on it.

[Bill-Gray]

You are correct. Compiled speed.c without wchar.h on X11, VT, DOS, DOSVGA, SDL1, SDL2, and WinGUI, and everything worked. I've removed that line (commit 348d53f).

[staalmannen]

Ok great! I will just import that one so that there are no merge conflicts if you decide to take this pull request.

[staalmannen]

I like that idea in general. In any case this likely still needs some time to be in a state where I'd find it possible to pull (which is of course @Bill-Gray's decision in any case).

I'll drop some notes in the suggested file changes in the meantime.

The general approach looks fine. I suggest to first check if there's a possible CI build option for Plan9/pcc, otherwise this port may be broken now or at any time later.

Some documentation stuff (starting with comments, extending to the Makefile and C headers, ending in the README) should be updated before the port, too.

Question @staalmannen: did you recheck all the demos? Can you share screenshots?
Do you know if there was a PR to PDCurses before and if yes can reference it?

Yes. In fact I directly imported the Plan9-related files from that PR ( wmcbrine/PDCurses#93 ). I did not expect that the two had diverged so much. I will definitely have to fix up the code specifically for the Bill-Gray fork and test-compile the new demos.

As soon as I have access to my Plan9 VM again I will try it and let you know if it works.

[staalmannen]

for a CI build option for Plan9 I only know of this one:
https://sourcehut.org/blog/2020-05-11-sourcehut-plus-plan-9/

but that is on an alternative git hosting site ( https://sr.ht/ )

[GitMensch]

Hm, actually I think the PR upstream would benefit of most of these adjustments, too :-)

[staalmannen]

[GitMensch]

@staalmannen Thank you for the screenshot, looks interesting. I assume that is a CHTYPE32 build, correct? Does the CHTYPE64 one with its additional attributes look different or aren't those supported yet (in this case: do you plan to add them)?

[staalmannen]

@staalmannen Thank you for the screenshot, looks interesting. I assume that is a CHTYPE32 build, correct? Does the CHTYPE64 one with its additional attributes look different or aren't those supported yet (in this case: do you plan to add them)?

reading curses.h, CHTYPE_64 seems to be the default so most likely that. I did not define PDC_WIDE though.

[GitMensch]

pdcplan9.c should be integrated in pdcutil.c to match with the other ports.
Other than this the PR is fine (still up to @Bill-Gray to include or not).

[GitMensch]

@Bill-Gray Do you see any issues with this new port or can I pull it for upcoming 4.2 (and add the one-line entry to docs/HISTORY.md, a file I'd still like you to come back)?

[Bill-Gray]

Simon's good with this port, and I see no problems (must confess I've not looked all that hard, but it looks as if Simon has, so I'll go with that.) I may have to add a Plan9 system to my OS menagerie...

[GitMensch]

After this is settled @staalmannen can now go upstream with more points solved, this is good.
@Bill-Gray I guess you'll do the History.md entry (possibly with the other points we found [or: I was whining over] and possibly with the preparation of PDCursesX)?