Updated eCos uSTL 1.3 port for review

Updated eCos uSTL 1.3 port for review

[cetoni GmbH - Uwe Kindler]

Hello John,

An updated port of the uSTL library 1.3 is available for review here:

http://www.cetoni.de/ecos_ustl20090820.zip

The code and all test cases now compile successfully in all 
configurations. (debug, release, with and without file streams, with and 
without file I/O package).

Now the libray and all tests cases do not pull in any external 
references to __cxa_guard_release() or any other functions from 
libsupc++ anymore.

So the problem from 
http://ecos.sourceware.org/ml/ecos-devel/2009-08/msg00031.html is not 
solved but it is not related to uSTL package anymore.

Is this package o.k. for CVS import or would you like to see some 
further changes?

To keep the changes to the original uSTL code low the flag -Wundef was 
removed from compiler flags for this package to avoid the need to change 
all checks of this kind: #if MACRO... in uSTL source.

Due to this also an eCos application that is build against this library 
and that uses uSTL should remove -Wundef flag to avoid compiler 
warnings. Is this appropriate? If this is not appropriate then more 
changes to the original source would be necessary and a later reimport 
would be more complicated.

Regards, Uwe

Re: Updated eCos uSTL 1.3 port for review

[John Dallaway]

Hi Uwe

Uwe Kindler wrote:

> An updated port of the uSTL library 1.3 is available for review here:
> 
> http://www.cetoni.de/ecos_ustl20090820.zip

Great. I will take a look...

John Dallaway

Re: Updated eCos uSTL 1.3 port for review

[John Dallaway]

Hi Uwe

Uwe Kindler wrote:

> An updated port of the uSTL library 1.3 is available for review here:

[ snip ]

> Is this package o.k. for CVS import or would you like to see some
> further changes?
> 
> To keep the changes to the original uSTL code low the flag -Wundef was
> removed from compiler flags for this package to avoid the need to change
> all checks of this kind: #if MACRO... in uSTL source.
> 
> Due to this also an eCos application that is build against this library
> and that uses uSTL should remove -Wundef flag to avoid compiler
> warnings. Is this appropriate? If this is not appropriate then more
> changes to the original source would be necessary and a later reimport
> would be more complicated.

Looking at the diff against the master uSTL 1.3 sources I now think the
removal of -Wundef (as you are currently doing) is preferable. As things
are, modifications to the existing uSTL source files are very small.

I have set the default value of CYGCLS_USTL_FSTREAMS to be
CYGPKG_IO_FILEIO rather than 1. In other words, CYGCLS_USTL_FSTREAMS is
enabled by default only if CYGPKG_IO_FILEIO is loaded. This is more
convenient if you wish to add uSTL to a default eCos configuration and
don't need file I/O.

I note that the issues raised by Jifl have largely been addressed.
stdint.h remains in the uSTL package but it will be trivial to switch to
using a common stdint.h in the future simply by deleting the uSTL
version of the file. I will adjust the copyright years on the various
ECOSGPLCOPYRIGHT banners to reflect reality.

The package and tests build cleanly with -Wundef. I have run up a number
of the tests on x86 PC hardware without issues.

There is more work that could be done in adjusting to reduced eCos
configurations but the existing contribution is genuinely useful as-is
and the configury can be fine-tuned over time. I will check this in.

What are your plans for src/twothreads.cpp which is not built or
referenced anywhere? I will omit this for now.

Thank you for the contribution!

John Dallaway

Re: Updated eCos uSTL 1.3 port for review

[Uwe Kindler]

Hi John,

thank you for doing the review and for CVS import of the uSTL library. 
The file twothreads.cpp is not part of the patch and was included by 
mistake. So you can remove it.

Regards, Uwe