Stale timestamps with `build.py --watch`

[nexushoratio]

I run build.py with 'version_timestamp': True, and usually with --watch.

The other day I noticed that About IITC was showing a build time that was several days old. It was confusing because I was seeing code changes in the plugin I was working on. On closer inspection, I noticed that the // @version metadata and plugin_info.dateTimeVersion = in all userscripts were "wrong". Or at least not the values I expected. They would be correct on the first build after restarting build.py.

I just did some digging and saw that build_date and build_timestamp are only set once during the lifetime of the python process, in the call to settings.load().

Since I use ViolentMonkey as my userscript manager, and I happened to use it to explicitly load and track the plugin I was working on, it was loading that particular file fine. Only once in a very rare time do I want to load all of the rest of the files. To that extent, I would use VM's Update scripts feature. But, that would often fail because it would use the standard @version comparison and see the old versions. (Alternatively, I could manually force a load of specific user.js, but that requires many more clicks.)

Anyway, a simple change:

diff --git a/build.py b/build.py
index a02d77a6..c5a1b0c6 100755
--- a/build.py
+++ b/build.py
@@ -118,7 +118,7 @@ def watch(build_cb, *args, interval=1, **kwargs):
     Every found dependancy (to be watched for changes) build_cb must append
     to deps_list, which is passed to it as additional keyword argument.
     """
-    from time import ctime, sleep, time
+    from time import ctime, gmtime, sleep, strftime, time
     from traceback import print_exc

     basetime = None
@@ -134,6 +134,9 @@ def watch(build_cb, *args, interval=1, **kwargs):
             else:
                 continue
         basetime = time()
+        utc = gmtime(basetime)
+        settings.build_date = strftime('%Y-%m-%d-%H%M%S', utc)
+        settings.build_timestamp = strftime('%Y%m%d.%H%M%S', utc)
         print('\nrebuild started [{}]'.format(ctime(basetime)))
         watch_list = []
         try:

seems "good enough for me".

Is there anything wrong with that approach in a general case? As commented in another issue from 2019, around when this was implemented, this would cause userscript managers to reload the built files more often. At least for VM, this means once/day automatically, or upon explicit request (except in the case of the specific plugin I configure to track).

I don't know how it would affect other userscript managers. NB: Since I use VM on Firefox, I have to serve up the userscripts via a webserver rather than it tracking the files locally. Again, I have no idea what that experience would be like.