First version of AddressSanitizer docs; documentation for __has_feature(address_sanitizer)

[ramosian.glider, 2011-11-22 07:32:02]

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html
File docs/AddressSanitizer.html (right):

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:41: <h2 id="intro">Usage</h2>
Using ASan on Darwin requires disabling the address space layout randomization.
Should we mention it here or on our wikipage?

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:43: <tt>-faddress-sanitizer</tt> flag and
optimization level <tt>-O1</tt> or higher
Are there any obstacles for compiling with -O0? I think we'd better not lose
such an opportunity.

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:45: and exit.
Should we have a usage example here?

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:62: <ul> <li>Linux <ul> <li> i386 <li> x86_64 <li>
arm </ul>
s/arm/ARM (unless it's the value of the -arch flag)

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:70: <li> On 64-bit platforms AddressSanitizer uses
16+ Terabytes of virtual
"uses" -> "maps" or "may use up to"

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:72: <li> Static linking is not supported.
"...yet"?

[Evgeniy Stepanov, 2011-11-22 09:54:24]

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html
File docs/AddressSanitizer.html (right):

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:73: </ul>
Say something about building DSOs with ASan, and how you can not link it with a
non-instrumented binary. It's a common catch with python and java native
modules.

[kcc, 2011-11-22 19:20:59]

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html
File docs/AddressSanitizer.html (right):

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:41: <h2 id="intro">Usage</h2>
On 2011/11/22 07:32:02, ramosian.glider wrote:
> Using ASan on Darwin requires disabling the address space layout
randomization.
Next time. 
> Should we mention it here or on our wikipage?
Yes, please.

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:43: <tt>-faddress-sanitizer</tt> flag and
optimization level <tt>-O1</tt> or higher
On 2011/11/22 07:32:02, ramosian.glider wrote:
> Are there any obstacles for compiling with -O0? I think we'd better not lose
> such an opportunity.

This is the current state. Once fixed properly (as you are doing it) this will
be removed.

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:45: and exit.
On 2011/11/22 07:32:02, ramosian.glider wrote:
> Should we have a usage example here?
Next time.

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:62: <ul> <li>Linux <ul> <li> i386 <li> x86_64 <li>
arm </ul>
On 2011/11/22 07:32:02, ramosian.glider wrote:
> s/arm/ARM (unless it's the value of the -arch flag)

Done.

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:70: <li> On 64-bit platforms AddressSanitizer uses
16+ Terabytes of virtual
On 2011/11/22 07:32:02, ramosian.glider wrote:
> "uses" -> "maps" or "may use up to"
rephrased.

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:72: <li> Static linking is not supported.
On 2011/11/22 07:32:02, ramosian.glider wrote:
> "...yet"?
Mmm. I don't have plans to support it. So, don't want to promise it.

http://codereview.appspot.com/5430044/diff/1/docs/AddressSanitizer.html#newco...
docs/AddressSanitizer.html:73: </ul>
On 2011/11/22 09:54:25, Evgeniy Stepanov wrote:
> Say something about building DSOs with ASan, and how you can not link it with
a
> non-instrumented binary. It's a common catch with python and java native
> modules.

Can you phrase it? (not necessary for the first variant)