Re: eCos uSTL port patch for review

[John Dallaway <john at dallaway dot org dot uk>]

Hi Uwe (and Jifl)

Uwe Kindler wrote:

> to get this right - you would like to keep the original uSTL #if ...
> definitions in the uSTL source instead of my #if defined ... changes and
> to define the macros to 0 in config.h so that the changes to uSTL
> source are kept to a minimum and most of the changes are in eCos
> specific config.h file? But you would prefer to see my #if defined ...
> changes in the official uSTL tree of Mike Sharov?

Yes. The official uSTL tree would ideally build silently with "-Wundef"
but if that isn't going to happen then the "#define HAVE_INT64_T 0"
route seems better than making a lot of changes in the eCos version of
the uSTL sources.

>> I agree. Option 1 seems best. At some point in the future, C99
>> vsnprintf() behaviour can become the default.
> 
> O.k., i will provide a patch.

Thank you.

>> You may find that there are so many dependencies that this makes no
>> sense, but it is worth investigating.
> 
> I already tried this but it is hard to track all dependencies. Therefore
> I would propose the following strategy:
> 
> Step 1:
> If I implemented all of your change requests the uSTL library goes into
> the eCos CVS
> 
> Step 2:
> If the uSTL library is in the eCos CVS it can be optimized by me or by
> other users by making the package more configurable.

I'm OK with this approach. Adding the extra configury is an optimisation.

> A last question. The uSTL eCos package contains a stdint.h file because
> eCos libc does not provide one. Would you prefer to see this file in the
> libc library or should it stay in uSTL package.

stdint.h is a C99 header. Adding it seems harmless to me, but to what
extent should we continue to add C99 features piecemeal to the existing
eCos C library which is based on C89? Is there a specific version of
stdint.h that would be preferable to ensure maximum compatibility? I
would like to hear Jifl's opinion on this. Jifl?

John Dallaway