It sounds like the --ide=vs and bundle_data things are both quite new and not likely tested in conjunction, this doesn't sound all that surprising ...

though we should probably at least add some unit tests to check for this along with a fix.

bundle_data is an iOS / mac thing, no? Might just need to put it behind an if (mac || ios)

We could do that, or we could leave it generic and fix the ide generator. It kinda seems like we should fix the generator regardless. 

Regarding #3, I initially put them behind "if (is_ios || is_mac)" and was asked by my reviewers to move them out as this would make the BUILD.gn file easier to read (less conditional) since bundle_data is a no-op on all platform that do not use create_bundle (which is itself hidden in templates behind "if (is_ios)").

I think we should fix fix "gn gen --ide=vs" to support bundle_data since I plan to extend the ide generator to support Xcode and at that point I'll need to support bundle_data.

Yeah, I agree we should do the trivial fix to make this a NOP for the VS generator.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/a90c62f80a74f7abc647234f70ac596e86bf3dd2

commit a90c62f80a74f7abc647234f70ac596e86bf3dd2
Author: sdefresne <sdefresne@chromium.org>
Date: Sat Mar 19 13:34:28 2016

[GN] Fix "gn gen --ide=vs" to ignore bundle_data targets.

In order to keep the BUILD.gn more readable by avoiding conditionals,
Chromium defines bundle_data on every platforms (as in the absence of
create_bundle they are just no-op stamp targets).

Fix "gn gen --ide=vs" to ignore bundle_data targets but not the
create_bundle targets (as there should be no such targets on Windows).

BUG= 595915 

Review URL: https://codereview.chromium.org/1811203003

Cr-Commit-Position: refs/heads/master@{#382197}

[modify] https://crrev.com/a90c62f80a74f7abc647234f70ac596e86bf3dd2/tools/gn/visual_studio_writer.cc

I'm getting a failure on bundle_data for the Clank build.  Is it related to this bug or should it be filed separately?

[mfoltz@sebastian /usr/local/google/home/mfoltz/clank/src]$  gn args out/Default
Waiting for editor on "/media/ssd2/mfoltz/clank/src/out/Default/args.gn"...
Waiting for Emacs...
Generating files...
ERROR at //sql/BUILD.gn:72:1: Unknown function.
bundle_data("sql_unittests_bundle_data") {
^----------

args.gn:

target_os = "android"
target_cpu = "arm"  # (default)
is_debug = true  # (default)

# Other args you may want to set:
is_component_build = true
is_clang = true
symbol_level = 1  # Faster build with fewer symbols. -g1 rather than -g2
#enable_incremental_javac = true  # Much faster; experimental

@mfoltz - that looks like you have an out-of-date GN binary. Make sure you have run `gclient runhooks` recently; if the problem persists, delete the src/buildtools directory and re-run `gclient runhooks` again.

I'm now getting a different failure:
ERROR Visual Studio doesn't support 'test_support_bundle_data' target output type: Bundle data

Is this because it also needs to ignore test_support_bundle_data along with bundle_data? Given that we just had another regression after the previous fix, is there anyway we can have a simple test to ensure that gn gen --ide=vs doesn't break?

oh nvm, I see that test_support_bundle_data is still a "bundle_data" type. However depot_tools is up to date for me (latest revision is from Wed Mar 23 09:52:51 2016 +0000) and I'm still seeing this?

also my buildtools/win is from 3/14 so it has the latest change from Dirk (https://codereview.chromium.org/1786503003)

Re: c#10

That makes sense since I'm also hitting crbug.com/579176 which is preventing me from running gclient runhooks.  Thanks

The --ide=vs fix for bundle_data didn't land until 3/19 (https://codereview.chromium.org/1811203003) and hasn't been rolled into chromium yet. 

I can do a roll today.

ah, got it thanks. I thought it was already rolled in per 10.

This is rolled in now (assuming the roll sticks).