Revert "bug #33092 [DependencyInjection] Improve an exception message" #33108
Conversation
|
Let's not revert before we understand. |
f321d35 to
ed590ca
Compare
|
@fabpot a stacktrace from the original issue would be helpful :) there are a few possible known suspects: https://github.com/symfony/symfony/search?l=PHP&q=%22Class+%25s+used+for+service+%25s+cannot+be+found.%22 the rationale isnt 100% clear to me, i.e. why exactly didnt we detect Now, some passes will benefit from #33111 when running "before removing" (e.g. So we may have somewhat of a chicken/egg situation, nevertheless the behavior change is real. I guess the user may process such service IDs themselves (so we are as preservative as possible). For reproducible builds we don't involve class_exists (correct me if im wrong), as such we would make a big assumption on service IDs. Hence we provide an error as late as possible. |
I can answer that: because we do not want any normalization logic around ids. We got rid of any when making them case sensitive and that reduced the complexity by a nice margin. We do not want to introduce it again in a different flavor. |
|
sorry, ive updated my comment :) basically, why didnt we do #33092 when introducing ResolveClassPass |
|
the discussion im recalling is #28006, same topic really 😅 |
…ption message" (nicolas-grekas) This PR was merged into the 4.3 branch. Discussion ---------- Revert "bug #33092 [DependencyInjection] Improve an exception message" | Q | A | ------------- | --- | Branch? | 4.3 | Bug fix? | no | New feature? | no | BC breaks? | no | Deprecations? | no | Tests pass? | yes | Fixed tickets | - | License | MIT | Doc PR | - As reminded by @ro0NL in #33092 (comment), it looks like we forgot that `CheckDefinitionValidityPass` already checks and suggests for leading slashes. Why didn't you get the exception from `CheckDefinitionValidityPass` @fabpot? Commits ------- ed590ca Revert "bug #33092 [DependencyInjection] Improve an exception message (fabpot)"
As reminded by @ro0NL in #33092 (comment), it looks like we forgot that
CheckDefinitionValidityPassalready checks and suggests for leading slashes.Why didn't you get the exception from
CheckDefinitionValidityPass@fabpot?