Add GDB printers for several more vocabulary types in Chrome

Add GDB printers for several more vocabulary types in Chrome

I also refactored some of the structure and fixed a bug in RenderProcessHostImplPrinter when some debug information is missing.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=189875

[Jeffrey Yasskin, 2013-03-21 23:55:25]

Below is some sample output for most of the new printers. Feel free to ask for
different formatting.

(gdb) p entry.user_typed_url_
$7 = "https://developer.chrome.com/extensions/"
(gdb) p &entry.user_typed_url_
$8 = (GURL *) 0x6042002d0270

(gdb) p &search_text
$16 = (const string16 *) 0x60100008fb48
(gdb) p search_text
$17 = "sm"

(gdb) p &current_size_
$19 = (gfx::Size *) 0x604a0003c990
(gdb) p current_size_
$20 = 937x992

(gdb) p session_storage_namespace_
$22 = scoped_refptr((content::SessionStorageNamespaceImpl *)0x6006002b9870)
(gdb) p &session_storage_namespace_
$23 = (scoped_refptr<content::SessionStorageNamespaceImpl> *) 0x604a0003ccd8

  pending_and_current_web_ui_ = base::WeakPtr((content::WebUIImpl *)0x0),

(gdb) p accessibility_layout_callback_
$25 = ...
(gdb) p /r accessibility_layout_callback_
$26 = {
  <base::internal::CallbackBase> = {
    bind_state_ = {
      ptr_ = 0x600600403ed0
    }, 
    polymorphic_invoke_ = 0x7fffb8fcd160
     <base::internal::Invoker<0,
base::internal::BindState<base::internal::RunnableAdapter<void (*)()>, void (),
void ()>, void ()>::Run(base::internal::BindStateBase*)>
  }, <No data fields>}

(gdb) p ((base::internal::WeakReference::Flag
*)0x600e003bed10)->thread_checker_.lock_
$28 = Unlocked

(gdb) p hung_renderer_timer_.delay_
$31 = 0:00:30
(gdb) p /r hung_renderer_timer_.delay_
$32 = {
  delta_ = 30000000
}
(gdb) p &hung_renderer_timer_.delay_
$33 = (base::TimeDelta *) 0x604a0003cb18

(gdb) p hung_renderer_timer_.scheduled_run_time_  
$35 = 7 days, 5:37:58.027591
(gdb) p /r hung_renderer_timer_.scheduled_run_time_
$36 = {
  ticks_ = 625078027591
}

(gdb) p &registrar_
$48 = (content::NotificationRegistrar *) 0x604a0003b220
(gdb) p registrar_
$49 =
    Watching 1 notifications; print ((content::NotificationRegistrar
*)0x604a0003b220)->registered_ for details
(gdb) p registrar_.registered_
$50 = std::vector of length 1, capacity 1 = {{
    observer = 0x604a0003b198,
    type = 22,
    source = {
      ptr_ = 0x0
    }
  }}

[tony, 2013-03-22 00:05:04]

LGTM.  The output looks nice, if someone wants a different format, they can
change it and send it out for review :)

https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py
File tools/gdb/gdb_chrome.py (right):

https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py#ne...
tools/gdb/gdb_chrome.py:166: 'base::Time::kTimeTToMicrosecondsOffset')
Cute.

https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py#ne...
tools/gdb/gdb_chrome.py:169: int(val['us_'] - timet_offset)))
Does this subtraction work?  It looks like val['us_'] is a string?

[Jeffrey Yasskin, 2013-03-22 00:17:29]

https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py
File tools/gdb/gdb_chrome.py (right):

https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py#ne...
tools/gdb/gdb_chrome.py:120: if flag and flag.dereference()['is_valid_']:
Oh, one thing I wanted to ask: I could write this as either
flag.dereference()['is_valid_] or just flag['is_valid']. (Of all the things
gdb-python could have made implicit, they chose the difference between '.' and
'->'?!?) I think I slightly prefer the implicit form, but I can definitely see
an argument for being explicit. What do you think?

https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py#ne...
tools/gdb/gdb_chrome.py:169: int(val['us_'] - timet_offset)))
On 2013/03/22 00:05:05, tony wrote:
> Does this subtraction work?  It looks like val['us_'] is a string?

Yeah, 'val' is a base::Time, and val['us_'] means val.us_. Since 'us_' is an
int64 (http://cs/#chrome/src/base/time.h&l=478), gdb wraps it into something
that can subtract. I need to do the explicit int()  conversion on the result
because timedelta checks for an exact type, not just a duck type. Whee.

[tony, 2013-03-22 16:48:43]

still LGTM

https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py
File tools/gdb/gdb_chrome.py (right):

https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py#ne...
tools/gdb/gdb_chrome.py:120: if flag and flag.dereference()['is_valid_']:
On 2013/03/22 00:17:29, Jeffrey Yasskin wrote:
> Oh, one thing I wanted to ask: I could write this as either
> flag.dereference()['is_valid_] or just flag['is_valid']. (Of all the things
> gdb-python could have made implicit, they chose the difference between '.' and
> '->'?!?) I think I slightly prefer the implicit form, but I can definitely see
> an argument for being explicit. What do you think?

I would probably use the implicit flag['is_valid'], but either is OK with me.

https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py#ne...
tools/gdb/gdb_chrome.py:169: int(val['us_'] - timet_offset)))
On 2013/03/22 00:17:29, Jeffrey Yasskin wrote:
> On 2013/03/22 00:05:05, tony wrote:
> > Does this subtraction work?  It looks like val['us_'] is a string?
> 
> Yeah, 'val' is a base::Time, and val['us_'] means val.us_. Since 'us_' is an
> int64 (http://cs/#chrome/src/base/time.h&l=478), gdb wraps it into something
> that can subtract. I need to do the explicit int()  conversion on the result
> because timedelta checks for an exact type, not just a duck type. Whee.

I see. Thanks for explaining!

[Jeffrey Yasskin, 2013-03-22 19:28:52]

Committed patchset #2 manually as r189875 (presubmit successful).

[Jeffrey Yasskin, 2013-03-22 19:29:05]

On 2013/03/22 16:48:43, tony wrote:
> still LGTM
> 
> https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py
> File tools/gdb/gdb_chrome.py (right):
> 
>
https://codereview.chromium.org/12465020/diff/2001/tools/gdb/gdb_chrome.py#ne...
> tools/gdb/gdb_chrome.py:120: if flag and flag.dereference()['is_valid_']:
> On 2013/03/22 00:17:29, Jeffrey Yasskin wrote:
> > Oh, one thing I wanted to ask: I could write this as either
> > flag.dereference()['is_valid_] or just flag['is_valid']. (Of all the things
> > gdb-python could have made implicit, they chose the difference between '.'
and
> > '->'?!?) I think I slightly prefer the implicit form, but I can definitely
see
> > an argument for being explicit. What do you think?
> 
> I would probably use the implicit flag['is_valid'], but either is OK with me.

Done, thanks!