Pass target architecture to build/android/envsetup.sh

Pass target architecture to build/android/envsetup.sh

TARGET_PRODUCT and other parameters generated by lunch command is not
necessary any more since the build got switched to being based on
SDK/NDK. This CL lets the target architecture be passed from envsetup.sh.
Currently two arch's values are accepted: arm(default), x86.

TEST=Built successfully with both target architecture values - arm and x86
    using following commands:
    . build/android/envsetup.sh  -> arm
    . build/android/envsetup.sh -t arm -> arm
    . build/android/envsetup.sh -t x86 -> x86
    Other commands tested:
    . build/android/envsetup.sh -h -> prints help
    . build/android/envsetup.sh -t bogus -> prints help and stops

BUG=

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=165185

[Jinsuk Kim (do not use this), 2012-10-24 04:50:42]

This CL replaces https://chromiumcodereview.appspot.com/11186014/ Please take a
look. Thanks.

[Yaron, 2012-10-24 20:50:17]

In general seems good.

+Isaac to review bash scripts.

https://codereview.chromium.org/11260008/diff/1/build/android/envsetup_functi...
File build/android/envsetup_functions.sh (left):

https://codereview.chromium.org/11260008/diff/1/build/android/envsetup_functi...
build/android/envsetup_functions.sh:197: export
TARGET_PRODUCT="${TARGET_PRODUCT:-trygon}"
build/android/emulator.py refers to this environment variable. I'm not familiar
with the specifics of it, but it'll break on x86 without additional changes.

[Isaac (away), 2012-10-24 20:54:48]

Bash stuff looks generally good with some nits.  Yaron can give final look over.

https://codereview.chromium.org/11260008/diff/1/build/android/envsetup_functi...
File build/android/envsetup_functions.sh (right):

https://codereview.chromium.org/11260008/diff/1/build/android/envsetup_functi...
build/android/envsetup_functions.sh:157: usage() {
rename print_usage

https://codereview.chromium.org/11260008/diff/1/build/android/envsetup_functi...
build/android/envsetup_functions.sh:187: if [ $# -ge ${OPTIND} ]; then
I don't see where $OPTIND is defined.

[Jinsuk Kim (do not use this), 2012-10-25 01:03:59]

Thanks for the review. Please take another look.

https://codereview.chromium.org/11260008/diff/1/build/android/envsetup_functi...
File build/android/envsetup_functions.sh (left):

https://codereview.chromium.org/11260008/diff/1/build/android/envsetup_functi...
build/android/envsetup_functions.sh:197: export
TARGET_PRODUCT="${TARGET_PRODUCT:-trygon}"
Thanks for catching it. I reverted the change to have the TARGET_PRODUCT defined
as before.

On 2012/10/24 20:50:17, Yaron wrote:
> build/android/emulator.py refers to this environment variable. I'm not
familiar
> with the specifics of it, but it'll break on x86 without additional changes.

https://codereview.chromium.org/11260008/diff/1/build/android/envsetup_functi...
File build/android/envsetup_functions.sh (right):

https://codereview.chromium.org/11260008/diff/1/build/android/envsetup_functi...
build/android/envsetup_functions.sh:157: usage() {
On 2012/10/24 20:54:48, Isaac wrote:
> rename print_usage

Done.

https://codereview.chromium.org/11260008/diff/1/build/android/envsetup_functi...
build/android/envsetup_functions.sh:187: if [ $# -ge ${OPTIND} ]; then
On 2012/10/24 20:54:48, Isaac wrote:
> I don't see where $OPTIND is defined.

OPTIND (together with OPTERR, OPTARG) is defined (and dynamically updated) by
the function getopts to handled the currently parsed option and the associated
values. I added local statement at the top of the function to clarify they are
used in this method only.

[Isaac (away), 2012-10-25 01:10:46]

https://codereview.chromium.org/11260008/diff/6003/build/android/envsetup_fun...
File build/android/envsetup_functions.sh (right):

https://codereview.chromium.org/11260008/diff/6003/build/android/envsetup_fun...
build/android/envsetup_functions.sh:189: eval echo "Unexpected command line
argument: \${${OPTIND}}" >& 2
I'm having trouble figuring out what this is doing, can you explain when we need
eval?

[Jinsuk Kim (do not use this), 2012-10-25 01:25:26]

https://codereview.chromium.org/11260008/diff/6003/build/android/envsetup_fun...
File build/android/envsetup_functions.sh (right):

https://codereview.chromium.org/11260008/diff/6003/build/android/envsetup_fun...
build/android/envsetup_functions.sh:189: eval echo "Unexpected command line
argument: \${${OPTIND}}" >& 2
I wanted to print out the unexpected argument, but echo alone was not enough to
evaluate the expression ${5} (for instance, when OPTIND=5), for which I used
eval to evaluate it. My bash-fu is not good enough to come up with a simpler way
to do it otherwise...

On 2012/10/25 01:10:46, Isaac wrote:
> I'm having trouble figuring out what this is doing, can you explain when we
need
> eval?

[Isaac (away), 2012-10-25 07:44:08]

https://chromiumcodereview.appspot.com/11260008/diff/6003/build/android/envse...
File build/android/envsetup_functions.sh (right):

https://chromiumcodereview.appspot.com/11260008/diff/6003/build/android/envse...
build/android/envsetup_functions.sh:189: eval echo "Unexpected command line
argument: \${${OPTIND}}" >& 2
I think it might be better to emulate parse_args from
build/android/buildbot/buildbot_functions.sh.  To me, that one seems clearer. 
It uses shift and doesn't need lots of OPT variables.

[Jinsuk Kim (do not use this), 2012-10-25 07:58:46]

Initially I used a sample in a bash shell book about getopts and handled the
problems by looking at other code in the same directory (gdb_apk) once I found
that it is already being used. And now they look mostly the same. I don't feel
it is too much complicated a piece of code as it is now. Please confirm it
really needs rewriting. I'll look again. Thanks.

On 2012/10/25 07:44:08, Isaac wrote:
>
https://chromiumcodereview.appspot.com/11260008/diff/6003/build/android/envse...
> File build/android/envsetup_functions.sh (right):
> 
>
https://chromiumcodereview.appspot.com/11260008/diff/6003/build/android/envse...
> build/android/envsetup_functions.sh:189: eval echo "Unexpected command line
> argument: \${${OPTIND}}" >& 2
> I think it might be better to emulate parse_args from
> build/android/buildbot/buildbot_functions.sh.  To me, that one seems clearer. 
> It uses shift and doesn't need lots of OPT variables.

[Isaac (away), 2012-10-25 08:28:19]

I think it is pretty complicated but it is OK given it is also in gdb_apk

[bulach, 2012-10-25 08:35:01]

+torne, as this has also implications on his side.

[Torne, 2012-10-25 11:01:32]

On 2012/10/25 08:35:01, bulach wrote:
> +torne, as this has also implications on his side.

This shouldn't be a problem for WebView. Currently we only ever generate
makefiles for ARM. In future we will probably just generate multiple sets and
can pass the architecture in the same way as the SDK build.

So, no objections from me, carry on. :)

[Yaron, 2012-10-25 20:57:24]

lgtm

[Yaron, 2012-10-29 23:37:28]

On 2012/10/25 20:57:24, Yaron wrote:
> lgtm

This failed on the trybot:
http://build.chromium.org/p/tryserver.chromium/builders/android_dbg/builds/10...
so I couldn't land it. It looks like we do pass in lots of other arguments on
the bots

[Jinsuk Kim (do not use this), 2012-10-31 01:22:31]

Revised the way options are processed like Isaac suggested to handle the trybot
issue.

- Now unknown options don't stop the script but let it proceed to allow trybot
that comes with other options to work.  
- Changed the option names to be more descriptive (-t -> --target-arch, -h ->
--help)

On 2012/10/29 23:37:28, Yaron wrote:
> On 2012/10/25 20:57:24, Yaron wrote:
> > lgtm
> 
> This failed on the trybot:
>
http://build.chromium.org/p/tryserver.chromium/builders/android_dbg/builds/10...
> so I couldn't land it. It looks like we do pass in lots of other arguments on
> the bots

[Yaron, 2012-10-31 17:27:33]

landed as 165185