build_nexe.py: Allow --strip-debug for archives

build_nexe.py: Allow --strip-debug for archives

This will allow us to strip debug info from libpnacl_irt_shim.a
and reduce the chrome-packaging size from 400KB -> 100KB.

Error out when give --strip-all on archives.

Locally tested with:  https://codereview.chromium.org/11566017/

BUG=none

Committed: http://src.chromium.org/viewvc/native_client?view=rev&revision=10461

[Roland McGrath, 2012-12-13 20:51:39]

https://codereview.chromium.org/11576019/diff/1/build/build_nexe.py
File build/build_nexe.py (right):

https://codereview.chromium.org/11576019/diff/1/build/build_nexe.py#newcode499
build/build_nexe.py:499: # Also valid to strip archives, but strip_all seems
dangerous.
Indeed, --strip-all makes .a or .o files unusable.
--strip-debug is fine for them.

[Sam Clegg, 2012-12-13 20:55:58]

https://codereview.chromium.org/11576019/diff/1/build/build_nexe.py
File build/build_nexe.py (right):

https://codereview.chromium.org/11576019/diff/1/build/build_nexe.py#newcode500
build/build_nexe.py:500: if self.strip_debug:
This seems a little confusing to me.

If I specify strip-all on a archive I would expect it to at least strip debug
info, rather than do nothing.  But stripping everything and having it fail
wouldn't be so bad either since it would force me to do the right thing.

[jvoung - send to chromium..., 2012-12-13 21:23:22]

https://codereview.chromium.org/11576019/diff/1/build/build_nexe.py
File build/build_nexe.py (right):

https://codereview.chromium.org/11576019/diff/1/build/build_nexe.py#newcode500
build/build_nexe.py:500: if self.strip_debug:
On 2012/12/13 20:55:58, Sam Clegg wrote:
> This seems a little confusing to me.
> 
> If I specify strip-all on a archive I would expect it to at least strip debug
> info, rather than do nothing.  But stripping everything and having it fail
> wouldn't be so bad either since it would force me to do the right thing.


Yeah, it shouldn't do less that strip-debug...
Ok I'll pass strip-all straight through so that people can do the right thing in
the gyp files.

[noelallen1, 2012-12-14 00:20:38]

On archives, does the strip-all fail immediately, or when someone tries to
link/load it?   If it's immediately on the strip call then LGTM

[Roland McGrath, 2012-12-14 00:31:12]

On 2012/12/14 00:20:38, noelallen1 wrote:
> On archives, does the strip-all fail immediately, or when someone tries to
> link/load it?   If it's immediately on the strip call then LGTM

The strip succeeds (it's doing what the user asked).
It's just that what you're asking for is to make the file unusable for linking,
so you will get errors (probably just undefined symbols) at link time.

[noelallen_use_chromium, 2012-12-14 00:38:52]

Seems like we should error then.  Getting a message further down the pipe from
where the error actually occurred makes debugging harder.   Might as well error
out in build_nexe when you see it.

[jvoung (off chromium), 2012-12-14 01:27:48]

On 2012/12/14 00:38:52, noelallen wrote:
> Seems like we should error then.  Getting a message further down the pipe from
> where the error actually occurred makes debugging harder.   Might as well
error
> out in build_nexe when you see it.

Ok, done.

[noelallen_use_chromium, 2012-12-14 17:27:22]

lgtm

[commit-bot: I haz the power, 2012-12-14 17:44:55]

CQ is trying da patch. Follow status at
https://nativeclient-status.appspot.com/cq/jvoung@chromium.org/11576019/13004

[commit-bot: I haz the power, 2012-12-14 19:38:50]

Change committed as 10461