DCHECK earlier if pepper flash is missing even though FLAPPER_AVAILABLE is defined.

chrome/common/chrome_content_client.cc

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "chrome/common/chrome_content_client.h"
 
 #include "base/command_line.h"
 #include "base/file_util.h"
 #include "base/path_service.h"
 #include "base/process_util.h"
@@ skipping 213 matching lines @@
         switches::kEnableBundledPpapiFlash);
 #endif
     bundled_flapper_enabled &= !CommandLine::ForCurrentProcess()->HasSwitch(
                                    switches::kDisableBundledPpapiFlash);
     if (!bundled_flapper_enabled)
       return;
 
 #if defined(FLAPPER_AVAILABLE)
     if (!PathService::Get(chrome::FILE_PEPPER_FLASH_PLUGIN, &plugin.path))
       return;
+    DCHECK(file_util::PathExists(plugin.path))

[viettrungluu, 2012/04/18 16:24:39]

This, I believe, causes file IO to be done on the UI thread, which isn't good
practice even for a Debug build.

[Joao da Silva, 2012/04/19 13:27:22]

This function is similar to ComputeBuiltInPlugins() above, which also checks for
the existence of the plugins listed there (and that's why those don't break).
This is also invoked during startup and before thread restrictions are
activated. An asynchronous interface around this would be better, but isn't
worth the complication.

This is causing issues at other completely unrelated levels, and was quite a
pain to track down. Having an early fail here spares a lot of time, and only
causes the IO penalty on debug builds.

[viettrungluu, 2012/04/19 23:34:25]

ComputeBuiltInPlugins() should be fixed. Well, more generally, the
"path-getting" functions that check for existence of files should be fixed.
<shrug>
Which is unfortunate, since our startup time is getting more and more dismal.
And people wonder why.
Well, ok. Skip the "Use --..." bit of the message though. If someone is doing
Official Debug builds, then they should be able to figure it out.

[viettrungluu, 2012/04/20 05:32:11]

Forgot to say: lgtm with a less verbose message.

On 2012/04/19 23:34:25, viettrungluu wrote:

[Joao da Silva, 2012/04/20 10:00:30]

Removed the log message altogether, and left a comment instead. The purpose is
to point people affected by this in the right direction.
Yes indeed. I'm actually working on something related: http://crbug.com/119550

+        << "FLAPPER_AVAILABLE is defined but the plugin does not exist. "
+        << "Use --" << switches::kDisableBundledPpapiFlash;
     flash_version = FLAPPER_VERSION_STRING;
 #else
     LOG(ERROR) << "PPAPI Flash not included at build time.";
     return;
 #endif  // FLAPPER_AVAILABLE
   }
 
   std::vector<std::string> flash_version_numbers;
   base::SplitString(flash_version, '.', &flash_version_numbers);
   if (flash_version_numbers.size() < 1)
@@ skipping 209 matching lines @@
   DCHECK(sandbox_profile_resource_id);
   if (sandbox_type == CHROME_SANDBOX_TYPE_NACL_LOADER) {
     *sandbox_profile_resource_id = IDR_NACL_SANDBOX_PROFILE;
     return true;
   }
   return false;
 }
 #endif
 
 }  // namespace chrome