add zenity backend

[wheybags]

Hey, I'd like to use this and don't want to pull in all of gtk as a dependency, so I added a zenity backend, as you suggested in your readme.

[mlabbe]

This looks promising at a glance. When I have some free time I'll give it a closer look.

Currently, simple_exec.h has a comment saying it was taken from another repo (one of yours). By offering this PR, do you permanently and irrevocably agree to the code being licensed under the same license as the rest of NFD?

[wheybags]

Yup, sure. It is derived in part from an so post, there's another comment with a link in it to that too.

…

On Sep 17, 2017 10:39 PM, "Michael Labbe" ***@***.***> wrote: This looks promising at a glance. When I have some free time I'll give it a closer look. Currently, simple_exec.h has a comment saying it was taken from another repo (one of yours). By offering this PR, do you permanently and irrevocably agree to the code being licensed under the same license as the rest of NFD? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#34 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABCumkAxDBMU5Mj-pt6NBW1JURQChKyjks5sjYN1gaJpZM4PVL6E> .

[wheybags]

pushed a fix for trailing newlines in zenity output

[mlabbe]

Keep 'em coming. I have yet to look at these changes (busy life), but I will holistically look at them all as soon as I can.

[wheybags]

bump :v (also fixed another stupid issue, surprising what you miss before you try really using something :p)

[wheybags]

Thanks!

[mlabbe]

Almost there!

I'm noticing the following warnings popping up when building in release_x64 config with gcc 5.4.0:

../../src/simple_exec.h: In function ‘runCommandArray’:
../../src/simple_exec.h:99:13: warning: ignoring return value of ‘write’, declared with attribute warn_unused_result [-Wunused-result]
write(errPipe[WRITE_FD], &err, 1);
^
../../src/simple_exec.h:131:25: warning: ignoring return value of ‘read’, declared with attribute warn_unused_result [-Wunused-result]
read(errPipe[READ_FD], &errChar, 1);

We should check and correctly handle unexpected byte length reads. Are you up for doing a small PR that fixes these new warnings to the devel branch?

[wheybags]

These calls are just a simple 1-char write and read to indicate if the process was successfully execed, not successfully run, it could have exited with nonzero exit status, but this checks whether the call to execv succeeded, ie, does the binary we asked to run exist and is it marked executable. If the exec was successful the process will never get to the write, because it will never return from the exec call.

If the write fails... well shit, ain't much we can do, and the process is about to exit. The read failing is the normal course. We read into errChar, which is initialised to zero. If the exec succeeds, nothing gets written to that pipe, the child runs, exits and the pipe is automatically closed.
In this case, the read returns 0 bytes, and the destination buffer isn't modified, otherwise we write a 1 and that gets read. So, for the read, there's no need to use the return value.

I would happily make a pr to suppress the warnings though?

[mlabbe]

@wheybags Hello - this is the maintainer of Native FIle Dialogs. The Zenity backend appears to have a discrepancy with the other backends. Can you take a look at #95 and advise how to bring it in line with the other backends? Thanks.

[wheybags]

Sorry, having looked at the man page for zenity, I don't think there is any way to do that unfortunately.