On Thu, Aug 25, 2011 at 4:34 AM, Cyril Hrubis <[email protected]> wrote:
> Hi!
>> as bionic isn't as smart or big as glibc). I would also recommend you
>> to port "Busybox" for Android or download it from the Android market
>> itself. The "ash" shell available in Busybox should be enough for you
>> to get started with the scripts in LTP.
>
> If anything doesn't work with ash, just send a patch. There may be some
> bashishm lefovers that needs to be fixed. Ideally all the scripts should
> work with POSIX compilant shell.
>
>> There are some patches already available for porting LTP to Android :
>> http://comments.gmane.org/gmane.linux.ltp/14497
>> However, these patches may not be acceptable to the mainstream LTP as
>> it breaks some existing tests in the "common/normal" platforms. I
>> think we can go ahead with the way we support UCLINUX in LTP ..
>> ie: #ifdef UCLINUX
>> So, if we care to re-write those patches with #ifdef ANDROID (or
>> #ifndef ANDROID based on our logic) without breaking other tests, I
>> think we can officially welcome Android related patches to LTP.
>>
>> Cyril, Garrett, please let us know your thoughts on this.
>
> As noted by garret previously, do not remove references for config.h.
>
> Either use the configure script to create config.h or create config.h by
> hand (as written in INSTALL).
>
> Ideally most of the differencies should be detected with configure script
> (Writing rules for generating it is hard but we may help you).
>
> Do not write "removed XXX to do YYY" into comments to C sources. We have
> git to track the change descriptions. Ideally each logical set of
> changes should be in one patch. Please learn how to use git to create
> patches with reasonable description. You have a bigger chance to get the
> changes commited by small well described and reasonable bits.
>
> Do not send patches mixed with DOS CR+LF
>
> Most of the changes that patch are just lines commented out, that is
> hardly applicable, this needs to be addresed in some reasonable manner.
>
> Generally I'm not saying no. This would just need some more work done.
>
> And for the UCLINUX part, I'm not really happy that we have #ifdef
> UCLINUX spread around the codebase, but I could live with reasonable
> bits that needs to be there. The best I could think about is to create
> android-compat.h header to store all the android specific stuff. Which
> would be included in test.h if ANDROID is defined.
Missing tools are one thing. Mike F and I went through the
sourcebase a few years ago and cleaned out most of the bashisms, but
not everything was "fixed". And there are still some perl scripts
laying around that really could be converted to straight shell //
shell + awk + sed.
It would be nice to enumerate why defining UCLINUX=1 cannot
resolve all issues with Android out of the box.
Thanks,
-Garrett
------------------------------------------------------------------------------
EMC VNX: the world's simplest storage, starting under $10K
The only unified storage solution that offers unified management
Up to 160% more powerful than alternatives and 25% more efficient.
Guaranteed. http://p.sf.net/sfu/emc-vnx-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
