[sql] uClibc doesn't provide malloc_usable_size().

[sql] uClibc doesn't provide malloc_usable_size().

For a normal SQLite build using uClibc, autoconf would generate a
config.h which SQLite would use.  Chromium's build system doesn't make
it easy to use different versions of a file depending on platform.
Use SQLite's config.h support to build a hybrid which allows additional
preprocessor detection.

BUG=489784
Committed: https://crrev.com/f52071c0c0b28fc67315dc4d22b8119c10059f55
Cr-Commit-Position: refs/heads/master@{#361175}

[smcgruer2, 2015-11-20 14:46:16]

smcgruer@google.com changed reviewers:
+ smcgruer@google.com

[smcgruer2, 2015-11-20 14:46:17]

https://codereview.chromium.org/1437823005/diff/1/third_party/sqlite/amalgama...
File third_party/sqlite/amalgamation/config.h (right):

https://codereview.chromium.org/1437823005/diff/1/third_party/sqlite/amalgama...
third_party/sqlite/amalgamation/config.h:1: /* On Windows and OSX, SQLite uses
preprocessor macros to configure itself.  On
To get access to __GLIBC__ and __UCLIBC__ etc, one must #include <features.h>

[Scott Hess - ex-Googler, 2015-11-20 18:05:22]

features.h

[Scott Hess - ex-Googler, 2015-11-20 18:06:05]

Great!  Could you give this patch try?

https://codereview.chromium.org/1437823005/diff/1/third_party/sqlite/amalgama...
File third_party/sqlite/amalgamation/config.h (right):

https://codereview.chromium.org/1437823005/diff/1/third_party/sqlite/amalgama...
third_party/sqlite/amalgamation/config.h:1: /* On Windows and OSX, SQLite uses
preprocessor macros to configure itself.  On
On 2015/11/20 14:46:17, smcgruer wrote:
> To get access to __GLIBC__ and __UCLIBC__ etc, one must #include <features.h>

Acknowledged.

[Scott Hess - ex-Googler, 2015-11-20 18:06:13]

The CQ bit was checked by shess@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2015-11-20 18:07:32]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1437823005/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1437823005/20001

[Scott Hess - ex-Googler, 2015-11-20 18:43:05]

Comment features.h

[Scott Hess - ex-Googler, 2015-11-20 18:43:24]

Hmm.  I don't see Chromium using features.h, instead build_config.h has this:

#elif defined(__linux__)
#define OS_LINUX 1
// include a system header to pull in features.h for glibc/uclibc macros.
#include <unistd.h>
#if defined(__GLIBC__) && !defined(__UCLIBC__)
// we really are using glibc, not uClibc pretending to be glibc
#define LIBC_GLIBC 1
#endif

Do you think unistd.h would be a better choice?  I can't see why build_config.h
doesn't use features.h directly, possibly it is leftover from someone trying to
keep Linux and OSX in line.  features_test_macros(7) uses unistd.h as an
example, but I don't think it's saying you get features.h via unistd.h, I think
it's just an example of how macros in features.h affect unistd.h-defined
functionality.

I did find some random project including limits.h before testing __GLIBC__, for
portability reasons.

As things currently stand, this config.h is only used on actual Linux, since
Windows and OSX self-configure rather than using autoconf.  I think SQLite would
consider Android a Linux for these purposes, but thus far there is no need for
this on Android, so I'd prefer not to touch that build.  Other non-Linux POSIX
presumably either works already (and thus will continue to work), or nobody
cares.  So to some extent this is debating angels on a pinhead.

Bah.  I'm going to add a comment and leave it at that.  What good is owning a
directory if you can't just do stuff sometimes?

[smcgruer2, 2015-11-20 19:37:24]

At least on my machine I see:

$ grep -r features.h /usr/include/unistd.h 
#include <features.h>
$ grep -r "define.*__GLIBC__" /usr/include/features.h 
#define	__GLIBC__	2
$

Most of the system headers pull in features.h themself, so you can generally
rely on them if necessary. I'm not sure if features.h is actually a standard in
any way whatsoever.

[smcgruer2, 2015-11-20 19:43:36]

lgtm

[Scott Hess - ex-Googler, 2015-11-23 19:07:35]

On 2015/11/20 19:37:24, smcgruer wrote:
> At least on my machine I see:
> 
> $ grep -r features.h /usr/include/unistd.h 
> #include <features.h>
> $ grep -r "define.*__GLIBC__" /usr/include/features.h 
> #define	__GLIBC__	2
> $
> 
> Most of the system headers pull in features.h themself, so you can generally
> rely on them if necessary. I'm not sure if features.h is actually a standard
in
> any way whatsoever.

What I'm finding is that when people try to out-think things, they end up
including weirdo things with a comment about how they're really trying to get
features.h.  So, for now I'm just going to stick with features.h and a comment
about how to deal with failure :-).  Maybe random other events will clear things
up.

-scott

[Scott Hess - ex-Googler, 2015-11-23 19:08:27]

Description was changed from

==========
[sql] uClibc doesn't provide malloc_usable_size().

For a normal SQLite build using uClibc, autoconf would generate a
config.h which SQLite would use.  Chromium's build system doesn't make
it easy to use different versions of a file depending on platform.
Use SQLite's config.h support to build a hybrid which allows additional
preprocessor detection.

BUG=
==========

to

==========
[sql] uClibc doesn't provide malloc_usable_size().

For a normal SQLite build using uClibc, autoconf would generate a
config.h which SQLite would use.  Chromium's build system doesn't make
it easy to use different versions of a file depending on platform.
Use SQLite's config.h support to build a hybrid which allows additional
preprocessor detection.

BUG=489784
==========

[Scott Hess - ex-Googler, 2015-11-23 19:08:31]

The CQ bit was checked by shess@chromium.org

[commit-bot: I haz the power, 2015-11-23 19:09:10]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1437823005/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1437823005/40001

[commit-bot: I haz the power, 2015-11-23 20:39:36]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2015-11-23 20:40:38]

Patchset 3 (id:??) landed as
https://crrev.com/f52071c0c0b28fc67315dc4d22b8119c10059f55
Cr-Commit-Position: refs/heads/master@{#361175}