exec doesn't run on first iteration

[miseran]

Since the update to 1.10.0, the exec command and friends don't run on the very first iteration for me.

For example, with the following config file

conky.config = {
    background = false,
    out_to_console = true,
    out_to_x = false,
    update_interval = 1.0,
    total_run_times = 0,
};

conky.text = [[
==${exec echo exec}==${execi 5 echo execi}==${texeci 5 echo texeci}==
]];

I get the output

========
==exec==execi==texeci==
==exec==execi==texeci==
==exec==execi==texeci==

etc.

[willemw12]

Even worse, in cases where conky is only run once:

total_run_times = 1

(e.g. the display-loop is executed by a containing program), there is no output at all.

[ghost]

I can confirm this behaviour also for use-cases where conky's output is piped into a bar like dzen. There, when restarting/reloading conky, the old/original output of the {t}exec{i} sequences is still being displayed…

[madhat2r]

I am also having this issue...

[labichn]

I can also confirm this issue. This bug prevents a number of simple use cases: a volume bar popup, for example, should be a single execbar iteration (total_run_times=1) with a duration of a second or so (update_interval=1). To get the same behavior (because of this bug) you need a very tight update interval with a large number of iterations so the initial empty bar isn't visible. I dug into this a bit but lost track of the control-flow of the callback system.

[marcpayne]

The callback for exec (and others) is registered after all callbacks are run for the first time. In case anyone wants to investigate further, I'll try to explain what happens.

Starting from conky.cc, the following function calls take place:

update_text()
|
|--- generate_text()
     |
     |--- update_stuff() [defined in common.cc]
     |    |
     |    |--- conky::run_all_callbacks()
     |
     |--- generate_text_internal()
          |
          |--- (*obj->callbacks.print)(obj, p, p_max_size) [see print_exec() in exec.cc]
               |
               |--- conky::register_cb<exec_cb>(1, true, obj->data.s)

The ultimate purpose of conky::run_all_callbacks() is to ensure that the work() function of each registered callback is called. However, the first call to run_all_callbacks() happens before register_cb<exec_cb>, so there is no exec_cb object whose work() function we can call.

I think we need some serious refactoring to make this work as intended. It isn't straightforward to move the register_cb call somewhere more appropriate, and calling run_all_callbacks() later does not work, afaict.

[chepioq]

Hi marcpayne
Have you solve this problem ?

[marcpayne]

No, unfortunately. I recently started training for a new job and have not had time to work on this issue. I will put in some time during the holidays and try to come up with a solution.

[marcpayne]

I think I have a decent solution now. I decided to push the changes to my repo, but not submit a pull request yet. If anyone here is willing to test, please do and let me know how it works for you. I'll be happy to help if needed.

One thing I mentioned in my commit message is that texeci sometimes works on the first update, and sometimes doesn't. I assume that the thread doesn't always have time to complete on my system before the first set of results gets printed. Your results may vary. At least the non-threaded versions of exec all work as they should!

[lasers]

Closed by #192. Thanks. 👍