Re: [PATCH] Remove duplicate or commented-out #includes

[Pedro Alves <palves at redhat dot com>]

On 01/22/2019 04:07 AM, Tom Tromey wrote:
>>> It would be possible to sort #includes, or maybe run
>>> include-what-you-use on gdb, but I haven't tried that.  I'd be
>>> interested to hear if you think this would be worthwhile, thoug.
> 
> Gary> I'd consider it worthwhile.  I'd really like sorted #includes.
> Gary> Some files have random whitespace in the list too, that maybe
> Gary> made sense once.
> 
> I wrote a script to sort the includes.  The output is rather voluminous,
> not totally sure yet how I will submit it.
> 
> It doesn't try to do a 100% job.  For example, some files in gdb include
> other files at random spots -- not all includes are at the top of the
> file -- and the script doesn't try to handle this.
> 
> It did find 40 .h files that don't have include guards.  Maybe I will
> try to automatically fix these.
> 
> Building revealed a few minor order dependencies in the headers.  I'll
> submit this as an initial cleanup.
> 
> 
> I chose to have the includes ordered this way:
> 
> First stanza:
> 
> 1. defs.h (or server.h or common-defs.h)
> 2. for a .c file, the corresponding .h if it exists
>    (two exceptions were needed to this rule)
> 
> Second stanza (stanzas separated by a blank line) holds system headers.
> 
> Third stanzas holds includes of headers in binutils-gdb but not part of
> gdb proper.
> 
> Fourth stanza is gdb-specific headers.
> 
> 
> It's reasonably easy to change this around though.

That order seems fine to me.  Thanks a lot for doing that!  Looking forward.

Thanks,
Pedro Alves

> It would be possible to modify this same script to try removing includes
> from .c files and seeing whether the file still compiles.  It would just
> take a long time on this machine I have at the moment.
> 
> If anyone wants to try it out for themselves, I can send the
> instructions.