fix: build log_proxy command in wrapper as an argv array, don't use gshell parser

[mk-fg]

This should fix passing -T "(null)" to log_proxy introduced by #35 in a hopefully slightly more obvious/cleaner way, and fix #37.

g_spawn_command_line_async() was replaced by g_spawn_async(), so that instead of earlier workflow of:

• sprintf-template a command line of arguments.
• Run g_spawn_command_line_async on it.
• Which runs g_shell_parse_argv to parse command line of arguments to gchar **argv.
• Run g_spawn_async with this argv.

It's just "build argv, run it with g_spawn_async", which eliminates unnecessary template-and-parse-it-back round-trip, and fixes string-escaping issues introduced by it cleanly.

Alternative can be to build a dynamic string, using g_shell_quote where string arguments are passed-through, but that seems unnecessary when it's just immediately parsed back to argv array anyway, and more error-prone in general.

Also removed glib/* component includes at the top, as these are redundant with #include <glib.h> which loads all of them, and are insufficient/incorrect otherwise anyway.

[thebaptiste]

Many thanks for this nice improvement. I'm happy to remove my ugly patch about "(null)" (which I wasn't very proud of)

[mk-fg]

Thanks for merging.

Oh, and also, can you hold off release for one more PR?

While going over the code again, I've noticed that my slapdash implementation of timestamp prefix in #35 was hugging lock in a loop unnecessarily ("continue" on EAGAIN without releasing the lock), so I made a patch to make it simpler and fix the issue, but didn't test it, which I can probably do and submit as a PR in the next hour or two.

[mk-fg]

Specific fix that I meant above: https://github.com/mk-fg/log_proxy/commit/ed0e3d7

[mk-fg]

Tested it just now, seem to work with/without prefix, added PR as #40.
I think everything is freed properly, and it's actually much less changes and code from pre-PR-35 than #35 itself is, which I think is always a plus too.
util.c can probably benefit from a cleanup to use more glib primitives, and is kinda redundant with what glib itself is.

[thebaptiste]

Thanks for merging.

Oh, and also, can you hold off release for one more PR?

While going over the code again, I've noticed that my slapdash implementation of timestamp prefix in #35 was hugging lock in a loop unnecessarily ("continue" on EAGAIN without releasing the lock), so I made a patch to make it simpler and fix the issue, but didn't test it, which I can probably do and submit as a PR in the next hour or two.

Yes it's probably simpler, safer and faster to have only one write per line.