[PATCH] clang-tidy docs

Alexander Kornienko alexfh at google.com
Mon Jul 27 03:56:55 PDT 2015


Manuel, thanks for the correction. Omitting quotes would be a problem with
`-checks *`, not `-checks=*` (which is used in the docs).

Aaron, this patch looks almost good then. Note that the description of
command-line arguments is just a dump of `clang-tidy -help`, so you need to
fix the documentation in the code and then paste `clang-tidy -help` output
to the .rst file and indent it appropriately.

Could you also check whether the -config=... examples work on windows?

It might also be useful to add a section describing the windows-specific
aspects of clang-tidy usage some time in the future.

-- Alex

On Mon, Jul 27, 2015 at 11:57 AM, Manuel Klimek <klimek at google.com> wrote:

> I think that something starts with -check= on the disk is low probability
> enough that we don't lose much by not escaping it for unix users, while
> gaining a lot less confusion on windows.
>
> On Fri, Jul 24, 2015 at 5:39 PM Alexander Kornienko <alexfh at google.com>
> wrote:
>
>> I think, we need to leave the examples valid for unix-like shells and add
>> a short section describing differences of shells or giving windows-specific
>> usage instructions. Some examples are just impossible to make compatible
>> with all shells (e.g. -checks='*', even though this is not particularly
>> useful).
>>
>>
>> On Fri, Jul 24, 2015 at 4:52 PM, Aaron Ballman <aaron at aaronballman.com>
>> wrote:
>>
>>> On Fri, Jul 24, 2015 at 8:44 AM, Manuel Klimek <klimek at google.com>
>>> wrote:
>>> > On Fri, Jul 24, 2015 at 2:34 PM Aaron Ballman <aaron at aaronballman.com>
>>> > wrote:
>>> >>
>>> >> On Fri, Jul 24, 2015 at 8:32 AM, Manuel Klimek <klimek at google.com>
>>> wrote:
>>> >> > Seems like we need different instructions for different shells then
>>> :(
>>> >> > The problem is that otherwise the -*... can be subject to shell
>>> >> > expansion if
>>> >> > it happens to match some files.
>>> >>
>>> >> Ah, I kind of wondered if this was a shell issue. Thank you for the
>>> >> verification!
>>> >>
>>> >> Do you think it makes sense to update the option parsing code to strip
>>> >> the single quotes if they are present?
>>> >
>>> >
>>> > No, I don't think it's the tool's job to handle idiosyncrasies of the
>>> > various shells.
>>> > For the docs I see two possibilities:
>>> > a) have 2 versions, one for cmd.exe, one for *sh.
>>> > b) the probability that users will actually have file named
>>> -something, is
>>> > not that high, we use the non-quoted version
>>>
>>> I kind of lean towards (b) with the understanding (which may be
>>> incorrect) that users of the shell are expected to understand when to
>>> quote arguments and when not to. That being said, I don't have a
>>> strong opinion on it.
>>>
>>> ~Aaron
>>>
>>> >
>>> > Alex, thoughts?
>>> >
>>> >>
>>> >>
>>> >> ~Aaron
>>> >>
>>> >> >
>>> >> >
>>> >> > On Thu, Jul 23, 2015 at 7:49 PM Aaron Ballman <
>>> aaron at aaronballman.com>
>>> >> > wrote:
>>> >> >>
>>> >> >> This patch addresses two issues (I can split the patch if it is
>>> >> >> desired):
>>> >> >>
>>> >> >> 1) The docs have some non-ASCII characters in them that aren't
>>> really
>>> >> >> required.
>>> >> >> 2) The docs suggest setting the checks using single quotes, which
>>> does
>>> >> >> not work (at least, on Windows).
>>> >> >>
>>> >> >> When you specify checks like -checks='-*,misc-some-check', the
>>> single
>>> >> >> quotes are not stripped by the option parser. When converting the
>>> >> >> flags into globs to pass along to regex, the single quotes remain
>>> as
>>> >> >> part of the regular expression, and do not match appropriately.
>>> When
>>> >> >> the single quotes are left off, the globs are correctly generated.
>>> >> >>
>>> >> >> ~Aaron
>>>
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20150727/7614e0d1/attachment.html>


More information about the cfe-commits mailing list