fix(view): error on missing version

[wraithgar]

This fixes an error in npm show. When calling npm show with a specific
version of a package that does not exist, it does not show anything and
gives a zero exit code. This has been changed: now it gives a 404 Error
similar to if the package does not exist. Can be tested with npm show
[email protected] (local: node bin/npm-cli.js info [email protected])

Fixes #4964

Credit: @lukaskuhn-lku

[wraithgar]

This is a continuation of #5011, I fixed the tests.

[npm-cli-bot]

no statistically significant performance changes detected

timing results

app-large	clean	lock-only	cache-only	cache-only peer-deps	modules-only	no-lock	no-cache	no-modules	no-clean	no-clean audit
npm@8	39.943 ±0.08	21.314 ±0.29	19.670 ±0.00	22.354 ±0.73	3.307 ±0.01	3.318 ±0.08	2.720 ±0.03	13.198 ±0.10	2.640 ±0.05	3.792 ±0.04
#5035	42.364 ±0.04	21.445 ±0.38	18.401 ±0.15	21.432 ±0.31	3.240 ±0.03	3.266 ±0.07	2.587 ±0.01	12.577 ±0.17	2.574 ±0.01	3.770 ±0.10

app-medium	clean	lock-only	cache-only	cache-only peer-deps	modules-only	no-lock	no-cache	no-modules	no-clean	no-clean audit
npm@8	30.172 ±0.95	15.573 ±0.11	13.788 ±0.18	14.835 ±0.26	3.027 ±0.01	3.136 ±0.07	2.644 ±0.03	9.374 ±0.09	2.497 ±0.04	3.425 ±0.01
#5035	29.501 ±0.24	15.553 ±0.06	13.798 ±0.04	15.021 ±0.39	2.999 ±0.05	2.971 ±0.05	2.653 ±0.03	9.268 ±0.03	2.522 ±0.04	3.383 ±0.02

[ljharb]

What about a dist-tag that doesn’t exist?

[wraithgar]

What about a dist-tag that doesn’t exist?

This PR also fixes that situation

~/D/n/cli (gar/5011|✔) $ node . view lodash@asdf
npm ERR! code E404
npm ERR! 404 No match found for version asdf
npm ERR! 404 
npm ERR! 404  'lodash@asdf' is not in this registry.

[MarshallOfSound]

@wraithgar Probably not worth a new issue but I wanted to note that this caused a catastrophic relatively bad release infrastructure failure for the Electron project.

We have since recovered and worked around the issue and ensured the npm is correctly pinned in our deployment environment but this was IMO a drastic breaking change in behaviour in a patch bump of the NPM cli.

Specifically we were checking if a version had been published to npm via the CLI using:

npm show [email protected] --json

And assuming that "no output" meant the version had not been published and "valid output" could be parsed into the registry metadata. In both cases since the dawn of npm the command exited cleanly, this change caused a non-zero exit code in the unpublished scenario which then propagated as an error through our release scripts causing an infinite backoff-and-retry (as we assumed a non-zero exit code meant npm was having issues).

Not sure if this is even gonna scratch on the radar, but I'd consider a change in exit code semantics for a CLI tool a breaking change and would be reserved for the next major version of npm.