Closed xaqrox closed 3 years ago
reynoldsalec
commented
3 years ago @xaqrox I haven't been able to reproduce this on my machine (similar specs), is this running off a boilerplate repo that you can share with me? I could spin it up in P.sh and see if I can replicate.
xaqrox
commented
3 years ago @reynoldsalec sorry, it's a custom site for a client and i don't think i can share the repo. but i do have some news on this.
i tried pulling out as much customization as i could out of the .lando and .platform.app .ymls, to try to rule out something I was doing. but the problem was still happening.
the error message above is frustratingly empty, but fortunately i discovered "silly" verbosity, which finally revealed that the problem script was /helpers/psh-build.sh, so i looked at that script. i build lando 3.5.1 from source, and removed the set -e from that script, hoping to get more informative errors. to my surprise, that actually let the build proceed as normal, and i just performed a successful pull via platform.sh. so something seems to be killing that script that doesn't actually need to be fatal when set -e is on, but i'm not sure what that is.
for now i have my app working and i need to get some stuff done so i'm going to stop debugging this, but i'll post here if i have any more clues in the next couple weeks.
reynoldsalec
commented
3 years ago Huh, that's interesting @xaqrox, wish you had got an error, could help figure out the culprit.
For now I'm going to close this out unless you have further info, thanks for following up!
xaqrox
commented
2 years ago @reynoldsalec One more detail I'm hoping will be helpful. I was inspired to try removing the set -e by a similar issue with nvm: https://github.com/nvm-sh/nvm/issues/1985 Long story short, it seems nvm does some error handling that expects set -e to not be present, perhaps a similar thing was happening with the psh cli installer, it was expecting to gracefully handle some error that set -e just took and ran with. 🤷
reynoldsalec
commented
2 years ago That's super interesting @xaqrox, my guess is that we're in a weird middle ground where set -e is probably the best default, but that something(s) that throw an error during the process don't really matter and shouldn't stop the show.
@pirog guessing we can't safely get rid of set -e for all cases?
xaqrox
commented
2 years ago I finally figured this out. There was some bad stuff in my .environment file, which does get included, eventually, via the platformsh plugin's psh-build.sh script. But there were various levels of error suppression, so I didn't realize that was where the problem was until I added some debugging to psh-build.sh (with very helpful tips from @pirog) to figure out exactly where it was choking, and followed it back to .environment. Whew!
xaqrox
commented
2 years ago I am running into this again, leaving a note for myself to the thread where pirog taught me how tell lando to use a different plugin location: https://devwithlando.slack.com/archives/C2XBSHX8R/p1642093821085800
reynoldsalec
commented
2 years ago @xaqrox note that there are now "custom installation" instructions on every Lando plugin in their docs, so hopefully that's a good reminder as well on how to download a specific version of the plugin and play with it: https://docs.lando.dev/platformsh/getting-started.html#custom-installation
I'm on v3.5.1 and bundled Docker, on macOS 11.6. Here's the error:
And my .lando.yml:
The app starts up well enough to serves page but when I try to run
lando pullI get (sorry for -vvv output, error is at the bottom):