Unable to look up source error

[3590Duncan]

When I click on a file in the coverage report I get an error indicating

Error: Unable to lookup source: /var/www/ui/components/booking/agenda/ActivityContainer.jsx(ENOENT: no such file or directory

We use docker to run our application locally. I am wondering if I need to mount the coverage or NYC folders in order for this data to be accessible?
I have attached a screenshot if the output

[bahmutov]

yeah, you probably just mount the tests, right? So the docker container does not have access to the `ui/components` folder

…

On Wed, Mar 4, 2020 at 12:22 PM Duncan Anderson ***@***.***> wrote: When I click on a file in the coverage report I get an error indicating Error: Unable to lookup source: /var/www/ui/components/booking/agenda/ActivityContainer.jsx(ENOENT: no such file or directory We use docker to run our application locally. I am wondering if I need to mount the coverage or NYC folders in order for this data to be accessible? I have attached a screenshot if the output [image: Screen Shot 2020-03-04 at 9 21 26 AM] <https://user-images.githubusercontent.com/49245425/75905340-8baf9000-5df9-11ea-920a-bd4fa28c18e9.png> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#147?email_source=notifications&email_token=AAQ4BJXMMZBQMVWVUKAQUO3RF2E4DA5CNFSM4LBODLF2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4ISO4W5Q>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAQ4BJWAMPIQKMF6CPQO56TRF2E4DANCNFSM4LBODLFQ> .

-- Dr. Gleb Bahmutov, PhD Schedule video chat / phone call / meeting with me via https://calendly.com/bahmutov gleb.bahmutov@gmail.com @bahmutov <https://twitter.com/@bahmutov> https://glebbahmutov.com/ https://glebbahmutov.com/blog https://github.com/bahmutov

[josh-byster]

I have the same issue—right now I'm running:

docker-compose run -v "$(pwd)"/frontend:/var/temp prod-frontend /var/www/app/node_modules/.bin/nyc report --report-dir /var/temp/coverage --temp-dir /var/temp/.nyc_output --reporter=lcov --reporter=clover --reporter=json

Which is essentially running a container instance and then generating a coverage report in the container which then because it's mounted then gets transferred to the host. This is a workaround that works for me right now.

But the problem isn't that we're just mounting the tests. The problem (which I'm assuming is the same as @3590Duncan) is that the generated coverage report has paths like /var/www/ui/components but when we try to access those coverage reports locally outside the Docker container, the files may be at a path like /Users/josh/Documents/project/ui/components. Hence there's a mismatch because it's looking for /var/www/ui/components but that's the path inside the Docker container. So it would be really nice if somehow the paths can be rewritten.

[bahmutov]

Idea: if the source paths in .nyc_output/out.json are all "wrong" and cannot be found, maybe they are just from Docker instrumentation and really are present locally and should be found (before generating the report). Need to try this.

[bahmutov]

🎉 This issue has been resolved in version 3.3.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[bahmutov]

Example in https://github.com/bahmutov/app-in-docker-coverage-example

[josh-byster]

This looks like it works on my end after running within the CI environment instead of the container (much easier) and testing it with my Dockerized app. Thank you for adding this feature!

[Quasarman]

Hello,

I think I am currently facing a similiar issue that is not covered by this fix. I am also running the web application inside a docker contianer therefore the paths to the respective source files is not the same. They get replaced by the according absolute paths as described here when running the e2e tests for the first time. But when running them for the second time with e.g., a different user-agent configuration to simulate mobile behaviour, and this new configuration introduces files not included in the first run, their path is still the same as the path inside the docker container, leading to the described error:

Error

.nyc_output/out.json paths correctly replaced (first run without user-agent configuration)

.nyc_output/out.json paths not correctly replaced (second run with user-agent configuration, file was not part of first run)

The overall coverage report is then unable to find any source files introduced by the second (mobile config) run. Despite this the coverage behaviour of soure files present during both runs is as expected. Figure below shows if there is a condition leading to a branch that checks mobile behaviour it gets correctly merged with the first run ($Platform indicates if run on mobile device):

This leads me to the question: Are the paths in the out.json only checked and replaced for the first time e.g., if the files does not previously exist? How could I overcome this issue? Since I really like to have the complete coverage (desktop and mobile) runs merged in a single output to analyze.

I had a look at the source code but I'm a bit lost in there...

Thanks for any help

[Quasarman]

Maybe this has to do with checking if all the files are missing?

[Quasarman]

Update: This also seems to happen wen having multiple test files not just running the tests twice...