Make the other Java modules debuggable

Make the other Java modules debuggable


BUG=
TEST=


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=141327

[haitao.feng, 2012-06-08 05:57:54]

Make developer's life easier to step in Java files in those modules.

I am not quite sure the debugging strategy of Android Content Shell. We could
not build the project with "make -j4 content_shell_apk BUILDTYPE=Debug" as we
need "--gc-sections" to link the libcontent_shell_content_view.so, even the
linking is correct, we could not run the application as the size is out of
memory.

Now we build it with -Os, and sometimes the line number is not correct when
using gdb. The ideal solution from my point is to replace -Os with -O0 in the
release build.

[John Grabowski, 2012-06-08 17:21:13]

LGTM

Do you think we should follow up and have the debug setting of gyp passed into
ant to turn this on/off explicitly?

Internally we don't build native code with -g since the binaries are ungodly
large and the compile/install/debug workflow is excessive.  I am open to option.

[haitao.feng, 2012-06-08 22:16:31]

For Java part, I do not think we need an option as the android.jar in the SDK
contains debugging information. I have verified that by using "javap -l "an
class in the android.jar"". I do not know whether Java debugger works in the
same way with remote gdb, that is, the debugging information is stripped in the
binary in the device while we still could debug it at the host side with the
un-stripped jar.

For native parts, what is your method to debug without "-g"? From my experience,
for "-Os", gdb sometimes gives wrong line number. I manually modified the .mk
file to use -O0 to solve the problem which is a pain. I wish we have
BUILDTYPE=Release-O0 to turn Os to O0.

On 2012/06/08 17:21:13, John Grabowski wrote:
> LGTM
> 
> Do you think we should follow up and have the debug setting of gyp passed into
> ant to turn this on/off explicitly?
> 
> Internally we don't build native code with -g since the binaries are ungodly
> large and the compile/install/debug workflow is excessive.  I am open to
option.

[John Grabowski, 2012-06-08 22:19:24]

On 2012/06/08 22:16:31, haitao.feng wrote:
> For Java part, I do not think we need an option as the android.jar in the SDK
> contains debugging information. I have verified that by using "javap -l "an
> class in the android.jar"". I do not know whether Java debugger works in the
> same way with remote gdb, that is, the debugging information is stripped in
the
> binary in the device while we still could debug it at the host side with the
> un-stripped jar.
> 
> For native parts, what is your method to debug without "-g"? From my
experience,
> for "-Os", gdb sometimes gives wrong line number. I manually modified the .mk
> file to use -O0 to solve the problem which is a pain. I wish we have
> BUILDTYPE=Release-O0 to turn Os to O0.
> 
> On 2012/06/08 17:21:13, John Grabowski wrote:
> > LGTM
> > 
> > Do you think we should follow up and have the debug setting of gyp passed
into
> > ant to turn this on/off explicitly?
> > 
> > Internally we don't build native code with -g since the binaries are ungodly
> > large and the compile/install/debug workflow is excessive.  I am open to
> option.

When we get the wrong line number, we just deal, or disassemble.  It's usually
pretty close.

[commit-bot: I haz the power, 2012-06-08 22:26:51]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/haitao.feng@intel.com/10541068/6

[commit-bot: I haz the power, 2012-06-08 23:45:08]

Change committed as 141327