Re: [PATCH REPOST] dtc: integrate gpp

On 22:11 Tue 21 Aug     , Stephen Warren wrote:

From: Stephen Warren <redacted>

I figured the easiest way to get named constants into dtc would be to
re-use an existing C pre-processor implemntation. Surprisingly, I could
not find many /good/ possibilities for this. I eventually found one called
gpp. This patch is an extremely quick-and-dirty integration of gpp into
dtc in order to solicit some feedback. What do people think of this
approach in general?

Issues:
* Fails some test-cases, related to dependency file generation, reading from
  stdin and include search paths.

TODO:
* Needs to be split into many smaller patches, such as:
  + Add unmodified upstream gpp.c.
  + Various different modifications to gpp.c to make it compile within dtc.
    Some of these might be able to go back upstream to gpp.
* Need a way to call gpp as a library, rather than faking a command-line for
  it. Perhaps this could go upstream too.
* Need to use a pipe instead of a temp file to pass the result from gpp to
  dtc. Or, if we rework gpp as a library, some intra-process way of streaming
  the output between processing steps. Perhaps this could go upstream too, if
  they accept the gpp-as-a-library concept.
* Need some better fixes for the const issues instead of xstrdup().
* Need to implement search paths for gpp includes.
* Probably more I forget; I originally prototyped this a while back.
* Test suite failure fixes, although at least some would be addressed by the
  points above.
---
Note: This patch is based on a slightly stale copy of dtc, plus some local
patches for other work I've sent, some of which has been accepted upstream
already, and some note. If this approach appeals to anyone, I will of course
rebase this.

I'm not sure it's rely the right way

I does the support but we may just use gcc for as it's done for linker script
but I will agree that with this dtc is standalone

Best Regards,