Added slim versions of output streams.

Added slim versions of output streams.

R=bmeurer@chromium.org, mstarzinger@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=22049

[Sven Panne, 2014-06-25 14:46:43]

No-bloat versions of the std::ostream stuff. Not used yet, I'd like to do that
in a separate CL, but can include that in this CL if this is preferred.

[Benedikt Meurer, 2014-06-25 14:48:16]

Please add tests.

https://codereview.chromium.org/352823003/diff/1/src/ostreams.h
File src/ostreams.h (right):

https://codereview.chromium.org/352823003/diff/1/src/ostreams.h#newcode101
src/ostreams.h:101: explicit OFStream(const char* p) : f_(fopen(p, "w")),
own_(true) { }
Please remove this constructor. It performs absolutely no error checking on the
result of fopen.

[Michael Starzinger, 2014-06-25 22:00:57]

Just one comment, rest looks OK to me, but I am no std expert at all.

https://codereview.chromium.org/352823003/diff/1/src/ostreams.cc
File src/ostreams.cc (right):

https://codereview.chromium.org/352823003/diff/1/src/ostreams.cc#newcode10
src/ostreams.cc:10: #define snprintf sprintf_s
Can we use SNPrintF from utils.h instead of using compiler-specific #ifdef's
here?

[Sven Panne, 2014-06-26 06:56:20]

I'll address the feedback, add the GYP/GN magic and even (*drumroll*) add a few
unit tests. Stay tuned...

https://codereview.chromium.org/352823003/diff/1/src/ostreams.cc
File src/ostreams.cc (right):

https://codereview.chromium.org/352823003/diff/1/src/ostreams.cc#newcode10
src/ostreams.cc:10: #define snprintf sprintf_s
On 2014/06/25 22:00:57, Michael Starzinger wrote:
> Can we use SNPrintF from utils.h instead of using compiler-specific #ifdef's
> here?

I intentionally avoided OS::SNPrintF because we increase dependecies for no good
reason. I think that src/base might be an even better place for ostreams.{cc,h},
so it's important to keep dependencies simple and don't introduce any cycles.

One could even go as far as using _MSC_VER directly, nuke the few V8_OVERRIDEs
and expand the (trivial) DISALLOW_COPY_AND_ASSIGNs by hand to get something
totally v8-independent, which would be nice (and is what I would have done
personally). But I wanted to avoid lengthy discussions... ;-)

https://codereview.chromium.org/352823003/diff/1/src/ostreams.h
File src/ostreams.h (right):

https://codereview.chromium.org/352823003/diff/1/src/ostreams.h#newcode101
src/ostreams.h:101: explicit OFStream(const char* p) : f_(fopen(p, "w")),
own_(true) { }
On 2014/06/25 14:48:16, Benedikt Meurer wrote:
> Please remove this constructor. It performs absolutely no error checking on
the
> result of fopen.

But this would be in line with ignoring basically all I/O errors in the rest of
v8... :-D  More seriously: Removing it is fine with me, but then I'll remove
close(), too, leaving just a plain decorator for FILE*.

Another approach might be introducing the 'failbit', but let's keep things
simple for now and only add things when we really need it.

[Sven Panne, 2014-06-26 12:11:23]

Addressed the feedback as discussed, overhauled the interface and added tests.
PTAL...

[Benedikt Meurer, 2014-06-26 15:38:50]

> https://codereview.chromium.org/352823003/diff/1/src/ostreams.h#newcode101
> src/ostreams.h:101: explicit OFStream(const char* p) : f_(fopen(p, "w")),
> own_(true) { }
> On 2014/06/25 14:48:16, Benedikt Meurer wrote:
> > Please remove this constructor. It performs absolutely no error checking on
> the
> > result of fopen.
> 
> But this would be in line with ignoring basically all I/O errors in the rest
of
> v8... :-D  More seriously: Removing it is fine with me, but then I'll remove
> close(), too, leaving just a plain decorator for FILE*.

That sounds like a plan. I was about to complain about close() too... :-)

[Benedikt Meurer, 2014-06-27 07:09:43]

LGTM

[Michael Starzinger, 2014-06-27 08:11:42]

LGTM.

[Sven Panne, 2014-06-27 08:42:31]

Committed patchset #3 manually as r22049 (presubmit successful).