[cfe-dev] LibTooling in the absence of a compilation database

[Manuel Klimek via cfe-dev]

On Wed, Dec 2, 2015 at 10:43 AM Russell Wallace <russell.wallace at gmail.com>
wrote:

> There are none that are guaranteed to work in all cases; nonetheless, if
> Clang, GCC or Microsoft C++ is presented with a 'hello world' test program
> with no other options, it will successfully compile it with its built-in
> defaults, so the defaults for a tool should be those that Clang would use
> in the same circumstances. (I do have some ideas for automated search for
> the correct options if the defaults fail, but in the meantime 'do what the
> actual compiler would do' would be a good baseline.)
>

The problem is that if this happens, it's hard to tell the user what they
need to do to get the tool to work (set up a compilation database).
How about we add to the error if there is not compilation database how to
run it in default mode (by adding -- at the end)?


>
> On Wed, Dec 2, 2015 at 9:39 AM, Manuel Klimek <klimek at google.com> wrote:
>
>> On Wed, Dec 2, 2015 at 10:35 AM Russell Wallace via cfe-dev <
>> cfe-dev at lists.llvm.org> wrote:
>>
>>> Tools built with LibTooling default to using a compilation database if
>>> one is available, which is good.
>>>
>>> In the case where a compilation database is not available, there is a
>>> problem: the default behavior is to exit with an error message, whereas in
>>> some cases the tool would be required to do the best it can anyway. This
>>> can be achieved by adding -- to the end of the command line, but this
>>> wouldn't be obvious to someone unfamiliar with Clang tools.
>>>
>>> One solution would be for the program to append -- to argv if it's not
>>> already present, but would this disable the default useful behavior of
>>> reading the compilation database if it is available?
>>>
>>> If so, is there a way for the program to say 'okay, check for a
>>> compilation database as normal, but if it's not present, let me know and
>>> I'll fill in some best-guess default options'?
>>>
>>
>> The problem is that for C++ there really are no best guess default
>> options. If you have ideas for what they should be, please make a proposal.
>>
>>
>>> _______________________________________________
>>> cfe-dev mailing list
>>> cfe-dev at lists.llvm.org
>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20151202/4df4f8e1/attachment.html>