[cmake-developers] Review request: Ninja-EXPORT_COMPILE_COMMANDS
Stephen Kelly
steveire at gmail.com
Fri May 11 05:58:41 EDT 2012
On 05/11/2012 11:55 AM, Erik Verbruggen wrote:
> Sorry for the late reply. Hopefully this makes it through the cmake list, of which I am not a member...
I'll forward just in case anyway.
>
> On May 10, 2012, at 10:45, Manuel Klimek wrote:
>
>> 3. A combination:
>> Add support to some target build systems (like ninja) to support the format.
>> Add support to meta build systems to output the format themselves for
>> targets that don't support it, and add implicit targets to reproduce
>> the information in the right spots for target build systems that
>> support it.
>>
>> I think (3) is the optimal solution - target build systems can opt in
>> to support the format, thus making use of their implicit knowledge
>> about the build, while meta-build systems can ensure that no target
>> system is left behind, using the common infrastructure to give basic
>> support with little code.
> Well, the first goal is to write tools to do things like refactoring on (larger) code-bases. For that, we "only" need to know how each file is compiled. Meaning the full command-line. We do not (yet?) need more. From my point of view, I would say that it is not important how we get that information, as long as we do get it. So, I would leave it to people writing (meta-)build systems to decide exactly how or where to emit this information. So yes, I am in favour of the "combination" approach.
>
> (Actually, I would not care if neither type of build systems would generate it. An IDE could do it, so could any user if he/she feels like it.)
>
>> From that perspective, I would argue that the proposed patch is a good
>> thing (as it is not a lot of code and supports a real current pain (at
>> least for me)), but we might want to rip it out once ninja supports
>> compile command output itself.
> Yes. Go for it.
>
> -- Erik.
Great, thanks,
More information about the cmake-developers
mailing list