Make creation of Puppeteer script part of reproduction step for autowebcompat - #6448
Make creation of Puppeteer script part of reproduction step for autowebcompat#6448ksy36 wants to merge 2 commits into
Conversation
c0b53c0 to
05a692a
Compare
ksy36
changed the title
bc015d4 to
2f565b7
Compare
2f565b7 to
bdb78c6
Compare
| agents/autowebcompat-repro/repro_reference.mjs \ | ||
| /app/repro/ | ||
| RUN cd /app/repro && PUPPETEER_SKIP_DOWNLOAD=1 npm ci --omit=dev | ||
|
|
There was a problem hiding this comment.
/app/repro contains node_modules, reference Puppeteer script and the script that agent writes (so it can access Puppeteer)
| @@ -0,0 +1,68 @@ | |||
| // REFERENCE — write your own script in this shape; don't copy this comment. | |||
| "Confirmation script did not reach a verdict in %s; " | ||
| "falling back to the reproduction steps", | ||
| channel, | ||
| ) |
There was a problem hiding this comment.
After a script has been created in BugReproduction, this part runs the script and if Firefox failure is confirmed, doesn't proceed to reproducing it using the agent for a given channel - sets result with reproduced: true, confirmed_by_script: true.
There was a problem hiding this comment.
This runs the script only in Firefox (basically what the agent has been doing as well)
| - If the reported broken behaviour reproduces in both browsers, it is not a | ||
| Firefox web-compat issue: set `failure_reason` to `non_compat`. | ||
| 4. If the issue reproduces AND the breakage is visual in nature (incorrect | ||
| 4. If the issue reproduces, write and run a Puppeteer script that drives the real |
There was a problem hiding this comment.
Hmm, so I guess a question is whether we want one script that launches both browsers, or a single script that can either launch Firefox or Chrome and produces different output (either exit code or stdout text) depending on whether the issue was reproduced or not. The advantage of the latter is that for future steps we want to be able to reuse the generated script where possible but not always require running in Chrome too (e.g. if we were using the script as the condition when running mozregression).
There was a problem hiding this comment.
Sounds good, I've changed it to launch one browser at a time and the script is checking expected behavior (and Firefox to fail and Chrome pass).
2 participants
No description provided.