Initial implementation of user-facing errors for state install and state uninstall.
#2877
Conversation
mitchell-as
force-pushed
the
mitchell/dx-2232
branch
3 times, most recently
from
0e31cd6
to
0786182
Compare
| type CommitError struct { | ||
| Type string | ||
| Message string | ||
| *locale.LocalizedError // for legacy, non-user-facing error usages |
There was a problem hiding this comment.
Since not all consumers are using user-facing errors, keep using the localized errors from the buildplanner for those consumers. The user-facing error rationalizer in this PR is making use of Type.
There was a problem hiding this comment.
I believe that for this to be used this type needs an Error() function that returns the localized error message. Unless there is another way to surface this that I am missing.
There was a problem hiding this comment.
I think it inherits locale.LocalizedError's Error() function. Otherwise there would be a compiler error that you cannot use a CommitError struct as an error type. For example, https://github.com/ActiveState/cli/pull/2877/files#diff-14612946db10f18744a68890b9e353cf87d4da9104592d9fe4346264ed83ef77R320 is the localized error used by non-user-facing errors applications.
mitchell-as
force-pushed
the
mitchell/dx-2232
branch
from
0786182
to
cb8520e
Compare
…`state uninstall`.
mitchell-as
force-pushed
the
mitchell/dx-2232
branch
from
cb8520e
to
52dc826
Compare
|
Test failures are not due to this PR. |
| // Error staging a commit during install. | ||
| case errors.As(*err, &commitError): | ||
| switch commitError.Type { | ||
| case bpModel.NotFoundErrorType: |
There was a problem hiding this comment.
I believe the Forbidden error type can also be covered by this case
There was a problem hiding this comment.
Not quite because of the message, but I've added another case with a slightly different message.
|
|
||
| // Error staging a commit during install. | ||
| case errors.As(*err, &commitError): | ||
| switch commitError.Type { |
There was a problem hiding this comment.
We should at least log the error message returned from the buildplanner for error types we are not handling here. We could also have a default case that just forwards the error message from the API. I know that for the ParseError we definitely want to do that this gives us the ability to have updated error messages in the case that the build expression format changes.
| type CommitError struct { | ||
| Type string | ||
| Message string | ||
| *locale.LocalizedError // for legacy, non-user-facing error usages |
There was a problem hiding this comment.
I believe that for this to be used this type needs an Error() function that returns the localized error message. Unless there is another way to surface this that I am missing.
| ) | ||
| case bpModel.NoChangeSinceLastCommitErrorType: | ||
| *err = errs.WrapUserFacing(*err, | ||
| locale.Tl("err_packages_exists", "That package is already installed."), |
There was a problem hiding this comment.
I think this message will have to be more generic as it could also arise when running state uninstall.
There was a problem hiding this comment.
May I ask how? If the user specifies a requirement that is not already installed, they'll get a requirementNotFoundErr, which is handled below.
There was a problem hiding this comment.
Ahh I see, I missed that part. Thanks for pointing that out.
| type CommitError struct { | ||
| Type string | ||
| Message string | ||
| *locale.LocalizedError // for legacy, non-user-facing error usages |
| ) | ||
| case bpModel.NoChangeSinceLastCommitErrorType: | ||
| *err = errs.WrapUserFacing(*err, | ||
| locale.Tl("err_packages_exists", "That package is already installed."), |
There was a problem hiding this comment.
Ahh I see, I missed that part. Thanks for pointing that out.
I don't really know what I'm doing. I tried to identify what could go wrong during install/uninstall, and tried some various install/uninstall scenarios to see what happened:
They should all return user-facing errors now.