-
Notifications
You must be signed in to change notification settings - Fork 1
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
"Unable to build typescript at rev 3d8cf62846c8bda828675239ddd9d4bd33e53d51" #41
Comments
Thanks for the report! I obviously couldn't test every single commit before publishing it 😄 Just to write down my thoughts, the unlink is a red herring; that path is to delete outputs if another error happens. I suspect that compiler error is the actual problem. How to fix it... usually this means that the code didn't build at the specific time, even though I ensure that I use a lockfile or
Just to clarify, what do you have some other thoughts here? Anything helps, unless you aren't meaning that. |
Well I peeked into the source code: This is the line it errored on: This was its Now maybe you have some more machinery in place than just trusting the |
I do, in that when there's no lockfile present, I grab the commit timestamp and then run |
Ahhh I see you mentioned Testing this specific commit isn't that important to me, you could always maintain a list of not working commits if you want? They'd of course have to be excluded from bisects but it'd be easier than making literally every single commit work. I only happened to stumble onto this commit because of bad luck, just happened to be picked during my bisect. I would say one improvement is that I had to reinstall Also I'd be willing to spin up a server to test every revision if you want. It could take several days to run to completion, I'd have to start the process of testing every revision before truly knowing an estimate, but I believe what I'd have to do would be |
If one commit fails, it's likely that other commits within the same timespan also fail, so I would try and aim for a more general fix. But, if it comes to it, I can of course figure out which range of commits need a specific version of the node types and check that when building.
That would be super interesting, but there are 35k+ commits on the repo... That's like 25 straight days of nonstop building assuming each takes a minute. I don't want to impose that cost on you, of course. Then again, the builds are mostly single threaded, so I bet one could install multiple instances of |
I've started processing the commits. This'll take several days to weeks at the current rate I'm seeing but it shouldn't cost me anything so don't worry about that. |
Update, I'm on revisions 20,000 out of 35,524. Currently 872 revisions have failed to build using |
I'm not surprised that so many don't work; the repo only started merge via squash a few years ago. |
Alright! It's done processing! The tail end was certainly quicker than the beginning. Here's some information I found interesting:
I still have the logs and timing information for every individual commit but I haven't uploaded those because they're a lot larger and mostly boring. I'd of course be willing to share them if you'd find that useful. I'm also willing to retest the 2,270 commits iteratively until you bring the number of broken commits down. HOWEVER I fully understand how much manual labor it could take to deal with these 2,270 commits. A simple approach to making |
Awesome! I really appreciate this data. I have some hope that I can fix some of these, but I suspect a bunch are just WIP commits from long ago (which a bisect with I can certainly include a list of all broken commits and error if any are checked out, though I'd like to take a crack at them regardless. |
I actually think they might be useful; if I'm going to fix certain classes of historical errors, it'd be nice to get them in buckets. E.g. the original report breaks due to:
So is a place where the repo was actually broken, and I could apply a patch for this kind of error by |
I also checked some of the ones that hang, and the first one I checked ( |
Hey! Thanks for the great tool! Saved me plenty of time already finding which PRs break my code.
Anyways, I was running a bisect and I got this error
Unable to build typescript at rev 3d8cf62846c8bda828675239ddd9d4bd33e53d51; please file a bug!
Just in case something funky was happening with how I'd set up my bisect, I stopped my bisect and then ranevery-ts switch 3d8cf62846c8bda828675239ddd9d4bd33e53d51
and sure enough I got the error again.Here's the full logs:
I think I have the general gist of the problem figured out but I'd be unsure how to fix it. Hopefully this is enough info for you!
The text was updated successfully, but these errors were encountered: