1487425 - AttributeError: 'dict' object has no attribute 'failed' when pushing / committing / ...

Status: RESOLVED FIXED

(Developer Infrastructure :: Lint and Formatting, defect)

Description

[:Gijs (he/him)]

full output:

$ hg push-to-try -m "Bug 1486824 - try: -b od -t none -p win64,macosx64,linux64,linux -u mochitest-bc --artifact "
Creating temporary commit for remote...
warning: no files linted
Error running mach:

    ['lint', '--quiet', '--workdir=staged']

The error occurred in code that was called by the mach command. This is either
a bug in the called code itself or in the way that mach is calling it.

You should consider filing a bug for this issue.

If filing a bug, please include the full output of mach, including this error
message.

The details of the failure are as follows:

AttributeError: 'dict' object has no attribute 'failed'

  File "tools/lint/mach_commands.py", line 44, in lint
    return cli.run(*runargs, **lintargs)
  File "python/mozlint/mozlint/cli.py", line 184, in run
    out = formatter(result).encode(sys.stdout.encoding or 'ascii', 'replace')
  File "python/mozlint/mozlint/formatters/stylish.py", line 55, in __call__
    failed = result.failed


This has been happening for a while now, and successive updates of m-c have somehow not fixed it, so here's a bug report...

Feels like the lack of files to lint means that the result object is empty and the formatter doesn't know how to deal with this.

Comment 1

[:Gijs (he/him)]

Oh, and I have:

pretxncommit.lint = python:path/to/m-c/tools/lint/hooks.py:hg


defined in .hg/hgrc on the trees where this happens. But happens on both my Windows and mac machines...

Comment 2

[Andrew Halberstadt [:ahal]]

Sorry about that, this is a regression from bug 1460856.

Comment 3

[Andrew Halberstadt [:ahal]]

Attachment: Bug 1487425 - [mozlint] Fix regression where 'roll' returns dict instead of ResultSummary when no files linted, r=gijs

This is a regression from bug 1460856.

Comment 4

[:Gijs (he/him)]

Comment on attachment 9005607 [details]
Bug 1487425 - [mozlint] Fix regression where 'roll' returns dict instead of ResultSummary when no files linted, r=gijs

:Gijs (he/him) has approved the revision.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/56d916d0bfb1
[mozlint] Fix regression where 'roll' returns dict instead of ResultSummary when no files linted, r=Gijs

Comment 6

[Andrew Halberstadt [:ahal]]

Somewhat related, we should try to get the `push-to-try` extension to skip the lint check when committing the temporary commit. There will never be a dirty working directory anyway (as |mach try| would have errored out if before this point if there were).

Comment 7

[Andrei Ciure[:aciure]]

https://hg.mozilla.org/mozilla-central/rev/56d916d0bfb1

Comment 8

[:Gijs (he/him)]

(In reply to Andrew Halberstadt [:ahal] from comment #6)
> Somewhat related, we should try to get the `push-to-try` extension to skip
> the lint check when committing the temporary commit. There will never be a
> dirty working directory anyway (as |mach try| would have errored out if
> before this point if there were).

I still use `hg push-to-try` directly, also with a dirty working dir (which then gets temp-committed). This is particularly useful for things like sec bugs, where I don't want to publish the original commit or bug number. Instead I `hg up` to a known good other bug, `hg diff -c <sec-bug-rev> | patch`, and then push-to-try .

Really what we need is a confidential copy of tryserver, but that's been on the wishlist for 5 years and has never happened, so I'm not holding my breath. In the meantime, if we could not break this type of usecase, that'd be great...

Comment 9

[Andrew Halberstadt [:ahal]]

(In reply to :Gijs (he/him) from comment #8)
> I still use `hg push-to-try` directly, also with a dirty working dir (which
> then gets temp-committed). This is particularly useful for things like sec
> bugs, where I don't want to publish the original commit or bug number.
> Instead I `hg up` to a known good other bug, `hg diff -c <sec-bug-rev> |
> patch`, and then push-to-try .
> 
> Really what we need is a confidential copy of tryserver, but that's been on
> the wishlist for 5 years and has never happened, so I'm not holding my
> breath. In the meantime, if we could not break this type of usecase, that'd
> be great...

Wfm, didn't realize anyone was using `hg push-to-try` directly. If it helps, we could add a --force flag to |mach try| to skip the uncommitted changes check. It's entirely superficial.