13 Jan 2020 |
oliver sanders | Just trying to fish out the line number | 00:14:50 |
oliver sanders | Will try with PYTHONASYNCIODEBUG=1 | 00:15:18 |
oliver sanders | * {
"errors": [
{
"message": "'name'",
"locations": [
{
"line": 2,
"column": 3
}
],
"path": [
"hold"
]
}
],
"data": {
"hold": null
}
}
| 00:15:46 |
oliver sanders | Nothing in the jupyterhub logging output, I would have expected something to get emitted from cylc-uiserver to stderr but I don't know... | 00:16:29 |
oliver sanders | Would be async code. | 00:16:34 |
oliver sanders | BTW: pdb breakpoints in cylc-uiserver surface in the jupyterhub console allowing singlesteping | 00:21:06 |
@kinow:matrix.org | In reply to @oliver-sanders2:matrix.org BTW: pdb breakpoints in cylc-uiserver surface in the jupyterhub console allowing singlesteping I'm not really familiar with pdb . But does that mean that it's possible to debug line-by-line of cylc-uiserver code? | 00:22:44 |
@kinow:matrix.org | In reply to @oliver-sanders2:matrix.org BTW: pdb breakpoints in cylc-uiserver surface in the jupyterhub console allowing singlesteping * I'm not really familiar with pdb . But is that good? Does that mean that it's possible to debug line-by-line of cylc-uiserver code? | 00:23:05 |
oliver sanders | Will probably work with whatever python debugger you use, pdb is a simple built-in one | 00:33:00 |
oliver sanders | Can debug line by line, frame by frame etc. | 00:33:10 |
Hilary Oliver | In reply to @oliver-sanders2:matrix.org BTW: pdb breakpoints in cylc-uiserver surface in the jupyterhub console allowing singlesteping Nice. Please note down for our workshop "how-to Cylc 8 debug" session! | 00:59:35 |
oliver sanders | It's a start but a pretty awful way to debug... | 01:10:12 |
Hilary Oliver | We all need to know how to effectively debug:
- cylc-flow back end (including code invoked in subprocesses)
- cylc-uiserver (including async routines)
- cylc-ui (JS and Vue, via browser dev tools)
- and the traffic in between components?
(debugging has got a lot more complicated in Cylc 8)
| 01:15:04 |
Hilary Oliver | In reply to @oliver-sanders2:matrix.org It's a start but a pretty awful way to debug... If we know the best way to debug (for each of the above cases) great; otherwise "it's a start" will have to do! | 01:15:34 |
David Sutherland | Oliver: that error points to something wrong in you request doc/mutation
On a separate but related note: I had to gerrymander the UIS to pass mutations to and receive response from WFSs | 01:26:50 |
oliver sanders | That error came from Resolvers.mutator | 01:30:33 |
oliver sanders | I'm starting to dig into that code now... | 01:31:16 |
oliver sanders | I've managed to shift that error which has now been replaced with list' object has no attribute 'values' | 01:31:54 |
oliver sanders | I've followed the execution right through to cylc.uiserver.resolvers.Resolvers.mutator:return but there it vanishes into async land. | 01:33:43 |
oliver sanders | * I've followed the execution right through to cylc/cylc/flow/network/schema.py(857)mutator():return but there it vanishes into async land. | 01:37:49 |
David Sutherland | it’s probably not processing the output correctly.. UIS is setup to deal with specifically the “command queued” type response from the WFS (cause I wanted the output to look the same as each other) | 02:05:06 |
oliver sanders | Yeah, I made WorkflowID (the input used for mutations) an input type with the fields name and user so I'm trying to refactor the resolver infrastructure to work with [{'name': 'foo', 'user': 'bar}, ...] rather than ['bar|foo', ...] . | 02:11:34 |
oliver sanders | Something's wrong with the processing/resolving. | 02:12:23 |
oliver sanders | Any tips for getting the full traceback | 02:13:28 |
oliver sanders | (or whether this change actually makes sense) | 02:13:45 |
David Sutherland | only at the UIS? try the workflow graphql | 02:13:55 |
oliver sanders | Only at the WFS, haven't needed to change anything UIS yet. | 02:17:22 |
oliver sanders | Have pushed my attempt thus far https://github.com/cylc/cylc-flow/pull/3469 | 02:17:44 |
Hilary Oliver | David Sutherland: I see from your email auto-reply that you're still on leave, till tomorrow? In which case, please don't feel obliged to chip in here all day today! | 02:18:42 |
David Sutherland | Are you using the CLI to hit the WFS graphql or the browser? | 02:19:40 |