[Toybox] [PATCH] Fix 32-bit bionic toybox build.

[enh]

On Sat, Jul 11, 2015 at 12:22 PM, Rob Landley <rob at landley.net> wrote:
> On 07/10/2015 09:25 PM, enh wrote:
>> (already submitted as https://android-review.googlesource.com/159035/
>> --- i was under pressure because i didn't notice the problem until i'd
>> already broken all the 32-bit builds. as usual i tested on 64-bit.
>> i've removed the non-standard basename_r and dirname_r from 64-bit
>
> Sigh. There's already a basename_r. Well of course there is.

it's worse than that --- there are several. the BSD one isn't the same
as the bionic one. (and neither of them are the same as yours.)

> It sounds like a good idea, personally I'd submit it to posix rather
> than remove it. A better fix would be to move basename_r() to
> portability.c and rely on libc to provide it if they have it. (I assume
> that your basename_r() and the implementation I did aren't too far apart
> behavior-wise?)

they're completely different. your basename_r is exactly the same as
the basename you get from bionic/glibc's <string.h> (as opposed to the
POSIX <libgen.h> one). the bionic basename_r is probably what you'd
expect if you wrote a potentially-modifying basename that used the
buffer it was passed. the BSD one is one you'd get if you tried that
but didn't have the sense to include the size_t.

> I need to dig up my AOSP build and try to chip the toolchain out to test
> build toybox. (Will static bionic builds run on ubuntu's kernel?)

afaik all the unit tests currently pass on Ubuntu 14.04.

>> bionic, but i can't fix the past. the build failure is because the
>> 32-bit bionic basename_r had the signature int basename_r(const char*,
>> char*, size_t). looking ahead maybe i should add a _BIONIC_NO_CRUFT so
>> non-historic 32-bit code can opt out of this kind of pollution.)
>
> Seriously, I'd just submit this one to the Austin group for the next
> posix/susvX spec. What the gnu guys did is a way bigger deviation than
> bionic here.

annoying though it is that they didn't give it a different name, i
think the GNU behavior's more generally useful.

and, like i said, LP64 bionic doesn't have basename_r.

>> commit f8b41e81fca2b53410f80c95f111c754d5eff99a
>> Author: Elliott Hughes <enh at google.com>
>> Date:   Fri Jul 10 19:11:18 2015 -0700
>
> I've applied this, but I'm open to using bionic's basename_r() instead
> of mine instead. FYI.
>
> Rob



-- 
Elliott Hughes - http://who/enh - http://jessies.org/~enh/
Android native code/tools questions? Mail me/drop by/add me as a reviewer.

 1436647399.0