Repository navigation
User-inputted module's main call called twice #293
Description
Opening up https://try.purescript.org/, I see that the module's main function is called twice, again. I thought this was fixed by #279, but maybe the recent es-module-shims update to 1.5.9 broke something?
Activity
andys8 commentedon Aug 10, 2022
In case the issue has its root cause in
es-module-shimsthere are two commits in recent releases that mention "initialization". Maybe interesting.guybedford/es-module-shims@1.5.9...1.5.12
Update: I don't think this will fix the issue.
I could reproduce the issue in Firefox, but not in Chrome. Could be browser or version related (e.g. one needs shims and the other doesn't).
https://discord.com/channels/864614189094928394/869607074152206397/1007075896005492837
andys8 commentedon Aug 11, 2022
I think I found the issue:
The docs of
es-module-shimshave this paragraph:This is not the case for try.purescript.org. Adding
import react from 'react';as the first import would indeed lead to a working example in both Chrome and Firefox.An alternative described in the docs is to activate
shimMode.JordanMartinez commentedon Aug 12, 2022
Yup, that's probably it.
pete-murphy commentedon Aug 27, 2022
Still seeing this issue in Firefox (104.0b10 (64-bit))
andys8 commentedon Aug 27, 2022
Cool, I see it both working (locally, 25a9095) and not working (try.purescript.org) at the same time using Firefox 103. Either there is a difference (not same code state, but I don't think so because the new comment is part of the html), frames behave differently on the domain compared to
localhostor its depending on (load) times or other external influence.screencast-2022-08-27_19.52.49.mp4
That makes it harder to reproduce locally, but the main issue is still the combination of iframes, settings listeners once, and
es-module-shimsintentionally crashing and executing scripts again. This combined with the fact thatrenderis usingsetInnerHTMLwhich doesn't set - but append - is brittle.Options
es-module-shimsor a larger change regarding listeners (once: true) and iframes<script type="module">import uuid from 'uuid';</script>). Without being able to reproduce locally, I don't know if this would fix therenderissue in FirefoxsetInnerHTMLand thereforerenderto be idempotent and set the content (and not append to it). This would change the behavior, in case you intentionally have multiplerendercalls. But if possible, this would be the easiest and likely (for the user) the most robust change.JordanMartinez commentedon Aug 27, 2022
For what it's worth, every time I've tried fixing this locally, it appeared to be fixed locally only for it to not be fixed on the actual site. Hence, it's hard to know when this problem has actually been solved.
andys8 commentedon Aug 28, 2022
Is the last suggested option viable?