guidosarducci:

Thanks for providing these details -- I'd heard of an issue but no one was providing specifics. The problem is that qos-scripts config parsing is fragile, and relies on some incorrect behaviour fixed in b6a1f43.

[[https://github.com//pull/1044|PR#1044]] fixes qos-scripts to be more robust and use documented config parsing, and works whether or not b6a1f43 is applied. Please try that and confirm.

lantsem:

I tried patch from PR#1044. qos-start seems to work ok with or without b6a1f43, and with default config or my custom one. But qos-stat still broken with b6a1f43. I forgot to write about it in original report, my mistake.

qos-stat

/usr/bin/qos-stat: eval: line 1: can't fork
/usr/bin/qos-stat: eval: line 1: can't fork
/usr/bin/qos-stat: eval: line 1: can't fork
/usr/bin/qos-stat: eval: line 1: can't fork
/usr/bin/qos-stat: eval: line 1: can't fork
/usr/bin/qos-stat: eval: line 1: can't fork
/usr/bin/qos-stat: eval: line 1: can't fork
/usr/bin/qos-stat: eval: line 1: can't fork
...

guidosarducci:

OK, thanks for confirming that my original fix worked.

I wasn't aware of qos-stat and don't use qos-scripts myself, so I appreciate your testing all this. I made a simple update to qos-stat in the PR, so please try [[https://github.com//pull/1044|PR#1044]] once more.

I'm also concluding most of the UCI callback usage in qos-scripts is not appropriate, and could be replaced with simper UCI reading. Would you mind testing another update with the standard/your custom config afterwards?

lantsem:

qos-stat is fixed now. Works with or without b6a1f43.

I can test future updates, thanks for your efforts!

guidosarducci:

Great, thanks for confirming that qos-stat works.

I've made one more update in [[https://github.com//pull/1044|PR#1044]] to remove unnecessary callbacks and simplify the config parsing. Would you mind reconfirming your earlier tests with qos-start and qos-stat? Thanks again.

lantsem:

Tested 2d007e1. I don't see any updates to qos-stat script, so old results are still relevant. As for generate.sh - I see problem with setting some values from config file. It sets classgroup and upload speed as default values from script (128 and Default), ignoring the given values in config file.

guidosarducci:

OK, I see the problem but didn't catch it because I'm testing with the default UCI config, which sets upload/classgroup to 128/Default. Thanks again for the catch.

It should be fixed now in the same PR.

lantsem:

PR#1044 a28baab: all good now, with or without b6a1f43.

guidosarducci:

Excellent news! The fix is now merged into the staging tree of @dedeckeh, so should appear in master sometime. Thanks for your help.