Experiments with SyncTeX

[lsoksane]

I have experimented a bit with SyncTeX between Textadept and Xreader from Linux Mint. The idea of SyncTeX is that you can jump from the TeX source code to the associated position in the pdf (forward search) and vice versa (backward search). Xreader has a D-Bus interface for this purpose, and I wrote two hackish Python scripts for the D-Bus communication.

I want textadept.run.run to do the following for a LaTeX buffer

1. Compile with the option synctex=1
2. Switch from the message buffer to the LaTeX buffer
3. Do forward search to sync with the pdf viewer
4. Start a Python server for backward search if needed

The code in textadept-synctex does this, except that instead of step 2 it simply closes the message buffer. The solution is far from ideal in many ways, but my main concern at the moment is that step 4 creates zombie processes. The server script exits immediately if there is already a server running, and the problem seems to be that os.spawn does not recognize that this child process is finished and should be terminated. This is what is see on Ubuntu 18.04

$ ps -ae | grep python
11690 pts/1 00:00:00 python
13001 ? 00:00:00 python <defunct>

with new defunct process created at each textadept.run.run after the first one. The zombies are killed when Textadept is closed. Also my other os.spawn calls create zombies unless I call wait() on the process, but wait() is not an option for the server process. Am I doing something wrong (apart from trying to pull off an ugly hack)?

[paaguti]

I exeprimented some time ago with the same configuration as you are using right now on a very old netbook. I'm a bit overwhelmed by work, but I could help you once I get back ro normal. I remember using epdfviewer and everything worked +/- flawlessly and with little or no Python code involved. If you have a sec, explore that path, pls. Look at my github https://github.com/paaguti for inspiration Best, /PA

…

On Wed, 27 Jan 2021 at 20:56, lsoksane ***@***.***> wrote: I have experimented a bit with SyncTeX between Textadept and Xreader from Linux Mint. The idea of SyncTeX is that you can jump from the TeX source code to the associated position in the pdf (forward search) and vice versa (backward search). Xreader has a D-Bus interface for this purpose, and I wrote two hackish Python scripts for the D-Bus communication. I want textadept.run.run to do the following for a LaTeX buffer 1. Compile with the option synctex=1 2. Switch from the message buffer to the LaTeX buffer 3. Do forward search to sync with the pdf viewer 4. Start a Python server for backward search if needed The code in textadept-synctex <https://github.com/lsoksane/textadept-synctex> does this, except that instead of step 2 it simply closes the message buffer. The solution is far from ideal in many ways, but my main concern at the moment is that step 4 creates zombie processes. The server script exits immediately if there is already a server running, and the problem seems to be that os.spawn does not recognize that this child process is finished and should be terminated. This is what is see on Ubuntu 18.04 $ ps -ae | grep python 11690 pts/1 00:00:00 python 13001 ? 00:00:00 python <defunct> with new defunct process created at each textadept.run.run after the first one. The zombies are killed when Textadept is closed. Also my other os.spawn calls create zombies unless I call wait() on the process, but wait() is not an option for the server process. Am I doing something wrong (apart from trying to pull off an ugly hack)? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#46>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAZKU4HB4E3CVQSLIW4XZZLS4BVX3ANCNFSM4WV27DYQ> .

-- Fragen sind nicht da um beantwortet zu werden, Fragen sind da um gestellt zu werden Georg Kreisler

[lsoksane]

Thanks a lot for the pointers. It seems to me that epdfviewer is not maintained anymore. Is this correct? I think that living with the D-Bus hassle of Xreader is a lesser evil.

[lsoksane]

I managed to solve the issue by forking the actual server startup (see the code in the linked repo). Then I can always call wait after os.spawn. I would still like to solve the minor issue with step 2, i.e. switching away from the message buffer rather than closing it.

[orbitalquark]

Normally you should be able to do so via `textadept.history.back()`[1], but this was buggy. I've pushed a fix[2] that will show up in tonight's nightly build. If you don't want to use history, you can try storing the LaTeX buffer somewhere and then use `view:goto_buffer()`[3]. [1]: https://orbitalquark.github.io/textadept/api.html#textadept.history.back [2]: 639c415 [3]: https://orbitalquark.github.io/textadept/api.html#view.goto_buffer

[lsoksane]

Thanks a lot. Using history works well. I will try to find time to make the code into a module. I have to learn how modules work first.

[paaguti]

Hi In an ideal world, I would have time to port it to GTK3. I try to compile it with every new Ubuntu and adapt it. Any mod will appear in my github. I don't know if that qualifies as 'maintenance' :-) Best, /PA