I realize I should have read #15 before, where people have apparently got it working with a different approach. So take this pull request with a grain of salt. Anyway, this fixes it for me. My problems were:

• gnuplot couldn't open any windows if called directly (without cmdproxy.exe, that is
• the value for comint-process-echoes was being incorrectly set in the *gnuplot* buffer, causing it to hang while in accept-process-output, called from comint-send-input.

Does anyone still working on this?

@bruceravel, could you give this a look over please? I can't test it on windows, but it does look good to me.

LGTM for me, but I don't have Windows and don't familiar Elisp on Windows.
@tarsius, any thoughts for this PR?

LGTM for me, but I don't have Windows and don't familiar Elisp on Windows.

I don't have Windows either but I maintain some popular packages that do things with paths and stuff, and that are being used by users of Windows...

And that experience tells me that whenever you do something to fix some Windows-specific breakage for one user, then you at the same time break it for another Windows user. That's because there are windows-nt and cygwin Emacsen and they are being used in combination with cygwin and/or non-cygwin tools. And cygwin can be configured in all kinds of crazy was and there are multiple cygwin forks with different. Also Windows now comes with a Linux implementation of sort, that's sure going to make it easier...

For Magit I have my "Windows guy" (who mostly knows what he is doing (I strongly suspect nobody fully knows what they are doing in this domain)) who handles these things for me, but for other packages I am opting for: "we can add your hack, but there must be an option so that users can enable it if only if it actually is broken for them without this".

And that experience tells me that whenever you do something to fix some Windows-specific breakage for one user, then you at the same time break it for another Windows user. That's because there are windows-nt and cygwin Emacsen and they are being used in combination with cygwin and/or non-cygwin tools. And cygwin can be configured in all kinds of crazy was and there are multiple cygwin forks with different. Also Windows now comes with a Linux implementation of sort, that's sure going to make it easier...

The code I added is only active when (eq window-system 'w32). Does cygwin Emacs respond to that?

Also, I believe that every package should behave more or less the way that Emacs itself handles windows stuff. So assume non-cygwin tools (the Gnuplot executable I was using was the "official one", if I remember correctly. Then give cygwin users a customization option. I.e. don't cater for specific packaging systems, that would be my advice.

That said, I'm not using windows anymore (yay!) and no Gnuplot either, so take this with a grain of salt and tweak the PR all you want!

My 2c

In my thoughts, remove all Windows support because I cannot test that environment.

we can add your hack, but there must be an option so that users can enable it if only if it actually is broken for them without this

I was just re-reading my code and it seems there is some kind of safety there. cygwin's shell-file-name shouldn't be cmdproxy.exe in principle. And you can also point gnuplot-program to whatever you want.

In my thoughts, remove all Windows support because I cannot test that environment.

That sounds a bit extreme. Emacs supports windows, so it's silly if one package purposedly doesn't support it. The lowest effort thing here is to add a new variable gnuplot-program-args, defaulting to nil and apply the make-comint function to it.

That way you don't break anything currently working, and you give poor Windows users a way out.

That sounds a bit extreme. Emacs supports windows, so it's silly if one package purposedly doesn't support it.

As GitHub Actions supports Windows, we could test on Windows.
BTW, what is the official Windows support? I know there is many difference cygwin and MSYS, but Emacs official page argued MSYS only. https://www.gnu.org/software/emacs/download.html#windows

The lowest effort thing here is to add a new variable gnuplot-program-args, defaulting to nil and apply the make-comint function to it.

I think this approach is better than now PR.

@joaotavora I followed your advice and implemented a gnuplot-program-args customizable variable. If that looks OK to you I can merge #65 and close this PR.

@joaotavora I followed your advice and implemented a gnuplot-program-args customizable variable. If that looks OK to you I can merge #65 and close this PR.

Have you tested on windows? I haven't and I won't since I don't use it anymore. Go and merge the other PR, I guess, but at least show the windows users in the README how to make use of that variable, or link them here at least.

I cannot test this on WIndows since I don't have a Windows machine. Setting the optional args functions correctly however.

Go and merge the other PR, I guess, but at least show the windows users in the README how to make use of that variable, or link them here at least.

Yes I intend to add an up-to-date section to the README.org before pushing by gathering info from this thread and #15. But I honestly don't know if Windows support will ever be a thing.

@joaotavora I added a Windows section on the README. Does it look OK to you?

The README isntructions look fine, though quite more complicated than what I had here, which seems harmless. If you know what you are doing, go for it!