Re: Android port of Emacs

[Óscar Fuentes]

Po Lu <luangruo@yahoo.com> writes:

> Óscar Fuentes <ofv@wanadoo.es> writes:
>
>> If I understand correctly, this means that you can't start Emacs from
>> Termux, but a specially-built Emacs could be launched as any other
>> Android app and then it could invoke Termux applications. Isn't the
>> "installed alongside Termux" part a bit complicated? I mean, if it is
>> not possible for Termux to build, package and install your port
>> alongside their other packages, how the user could do that?
>
> By configuring Emacs with:
>
>   ./configure --with-shared-user-id=com.termux
>
> I think.  Or perhaps org.termux?

Great.

>> How is that different from Termux themselves building and distributing
>> your port?
>
> Termux only packages command-line applications, while programs using GUI
> features can only be started by the system itself.
>
> HTH.

There is a Termux extension on the works for supporting X11 and Wayland,
so that can change on the future. Other projects (UserLand) provide a
graphical environment out of the box. I just tried Emacs on UserLand
(Emacs 28.2, the GTK build, from their Arch distribution), the
experience is not great, but it runs.

Installing Emacs as a separate step would not be an issue, but if the
user could launch your port from Termux/Userland or from elsewhere
having access to all those binaries, it would be a boom.

>> OTOH, I know from experience that creating, hosting and maintaining a
>> software distribution (and that would be what you are describing, IIUC)
>> is quite a lot of work, even when the number of packages is small-ish
>> (about 50-100 in my case, counting dependencies.) So having a
>> "hands-off" method for installing Emacs along with other software
>> packages is crucial for gaining a broad adoption.
>
> Well, if it helps, I'll see if I can scrape together some machines to
> distribute packages built this way.

Considering that we already have software distributions for Android,
that would be a waste of your time. And believe me, it is a lot of time,
building the packages is the easy part, the real work comes with the bug
reports.

Today I tried your port again for about an hour. It is obvious that you
use a hardware keyboard ;-) Emacs hides the on-screen keyboard too
often, even with touch-screen-display-keyboad set to `t': just try M-x
calc and do a simple operation. IMO Emacs should never hide the
on-screen keyboard, moreover when clicking with the finger over the
small clickable areas of menus and links is very difficult. And a quick
method to show the keyboard would be very welcome (having the volume
button trick to close Emacs is curious, I'm sure you have your reasons,
but if the keyboard is shown and the user can press C-g, isn't it better
than closing Emacs?)

Also, completion works weird: the sequence C-x b Mes TAB correctly
completes to *Messages*, but when ENTER is hit Emacs creates a buffer
names "Mes".

But those are minor issues that I'm sure will be resolved. I'm having
strong W32/Cygwin vibes. Nobody would say nowadays that the W32 port is
questionable because Cygwin exists. Likewise, this port shows lots of
potential as the right port for the general userbase that would like to
carry Emacs on our pockets.