Implement --omit, --raw_[context_]file=... for mksnapshot tool.

Implement --omit, --raw_[context_]file=... for mksnapshot tool.

These switches are analogous to --omit and --raw for js2c.py.
The goal it to allow embedders to externally load all (or generate) the V8 startup data.

There's a bit of gratuitous refactoring with the writing logic that helped me understand the code. Let me know if it's not an improvement.

BUG=355539
LOG=N
R=jochen@chromium.org, svenpanne@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=20941

[vogelheim, 2014-04-23 13:14:26]

Next steps would be:

1, add a variable to gyp files to generate raw files.
2, add logic in Chromium to (statically) use these.
3, put the runtime part of mksnapshot behind an API
   (essentially, rip out  SnapshotWriter), so this can
   (optionally) be done at runtime.

[jochen (gone - plz use gerrit), 2014-04-23 15:23:42]

lgtm

I guess make quickcheck still passes with this?

[Sven Panne, 2014-04-24 06:59:16]

https://codereview.chromium.org/249283002/diff/20001/src/flag-definitions.h
File src/flag-definitions.h (right):

https://codereview.chromium.org/249283002/diff/20001/src/flag-definitions.h#n...
src/flag-definitions.h:660: DEFINE_bool(omit, false, "Omit raw snapshot bytes in
generated code.")
These new flags don't make sense outside of mksnapshot, can we find a way to
restrict them to that? Having them e.g. in d8 is a bit confusing. I know this
problem has been there before, but now we have 4 confusing flags instead of 1.

https://codereview.chromium.org/249283002/diff/20001/src/mksnapshot.cc
File src/mksnapshot.cc (right):

https://codereview.chromium.org/249283002/diff/20001/src/mksnapshot.cc#newcode75
src/mksnapshot.cc:75: ~SnapshotWriter() {
Here and below: We normally use 1 blank line between function definitions in
classes and 2 blank lines between out-of-line function definitions. We sometimes
deviate from these rules for single one-line getter/setter functions.

https://codereview.chromium.org/249283002/diff/20001/src/mksnapshot.cc#newcod...
src/mksnapshot.cc:116: void WriteData(const char* prefix, const i::List<char>&
source_data,
Here and at other places: If parameters don't fit on a line, we usually have a
single parameter per line.

[jochen (gone - plz use gerrit), 2014-04-24 07:47:59]

On 2014/04/24 06:59:16, Sven Panne wrote:
> https://codereview.chromium.org/249283002/diff/20001/src/flag-definitions.h
> File src/flag-definitions.h (right):
> 
>
https://codereview.chromium.org/249283002/diff/20001/src/flag-definitions.h#n...
> src/flag-definitions.h:660: DEFINE_bool(omit, false, "Omit raw snapshot bytes
in
> generated code.")
> These new flags don't make sense outside of mksnapshot, can we find a way to
> restrict them to that? Having them e.g. in d8 is a bit confusing. I know this
> problem has been there before, but now we have 4 confusing flags instead of 1.
> 
> https://codereview.chromium.org/249283002/diff/20001/src/mksnapshot.cc
> File src/mksnapshot.cc (right):
> 
>
https://codereview.chromium.org/249283002/diff/20001/src/mksnapshot.cc#newcode75
> src/mksnapshot.cc:75: ~SnapshotWriter() {
> Here and below: We normally use 1 blank line between function definitions in
> classes and 2 blank lines between out-of-line function definitions. We
sometimes
> deviate from these rules for single one-line getter/setter functions.
> 
>
https://codereview.chromium.org/249283002/diff/20001/src/mksnapshot.cc#newcod...
> src/mksnapshot.cc:116: void WriteData(const char* prefix, const i::List<char>&
> source_data,
> Here and at other places: If parameters don't fit on a line, we usually have a
> single parameter per line.

I'm not really a huge fan of those unwritten coding style rules. The V8 style
guide says that the code should follow the Google C++ style guide :-/

[vogelheim, 2014-04-24 11:07:59]

PTAL. Not sure what to do abount the flags thing, though.

https://codereview.chromium.org/249283002/diff/20001/src/flag-definitions.h
File src/flag-definitions.h (right):

https://codereview.chromium.org/249283002/diff/20001/src/flag-definitions.h#n...
src/flag-definitions.h:660: DEFINE_bool(omit, false, "Omit raw snapshot bytes in
generated code.")
On 2014/04/24 06:59:17, Sven Panne wrote:
> These new flags don't make sense outside of mksnapshot, can we find a way to
> restrict them to that? Having them e.g. in d8 is a bit confusing. I know this
> problem has been there before, but now we have 4 confusing flags instead of 1.

I looked at this, thinking it should be really easy to just move those flags to
mksnapshot.cc which contains the main method. Except... the way the current
flags.h/.cc infrastructure is built, it relies on repeatedly including
flag-definitions.h with various defines set to 'extract' different data out of
the macros there. There appears to be no way to modify the data structures used
by flags.h/.cc other than flag-definitions.h; at least I didn't find any.

Honestly, I'm not super fond of this design, but I can't see any fix other than
refactoring the whole flags thing.

https://codereview.chromium.org/249283002/diff/20001/src/mksnapshot.cc
File src/mksnapshot.cc (right):

https://codereview.chromium.org/249283002/diff/20001/src/mksnapshot.cc#newcode75
src/mksnapshot.cc:75: ~SnapshotWriter() {
On 2014/04/24 06:59:17, Sven Panne wrote:
> Here and below: We normally use 1 blank line between function definitions in
> classes and 2 blank lines between out-of-line function definitions. We
sometimes
> deviate from these rules for single one-line getter/setter functions.

Done.

https://codereview.chromium.org/249283002/diff/20001/src/mksnapshot.cc#newcod...
src/mksnapshot.cc:116: void WriteData(const char* prefix, const i::List<char>&
source_data,
On 2014/04/24 06:59:17, Sven Panne wrote:
> Here and at other places: If parameters don't fit on a line, we usually have a
> single parameter per line.

Done.

[Sven Panne, 2014-04-24 11:12:43]

LGTM with a nit

https://codereview.chromium.org/249283002/diff/20001/src/flag-definitions.h
File src/flag-definitions.h (right):

https://codereview.chromium.org/249283002/diff/20001/src/flag-definitions.h#n...
src/flag-definitions.h:660: DEFINE_bool(omit, false, "Omit raw snapshot bytes in
generated code.")
On 2014/04/24 11:07:59, vogelheim wrote:
> [...] Honestly, I'm not super fond of this design, but I can't see any fix
other than
> refactoring the whole flags thing.

I suspected that... :-/ OK, then just add "(mksnapshot only)" to the
descriptions.

[vogelheim, 2014-04-24 11:24:15]

On 2014/04/24 11:12:43, Sven Panne wrote:
> LGTM with a nit
> 
> https://codereview.chromium.org/249283002/diff/20001/src/flag-definitions.h
> File src/flag-definitions.h (right):
> 
>
https://codereview.chromium.org/249283002/diff/20001/src/flag-definitions.h#n...
> src/flag-definitions.h:660: DEFINE_bool(omit, false, "Omit raw snapshot bytes
in
> generated code.")
> On 2014/04/24 11:07:59, vogelheim wrote:
> > [...] Honestly, I'm not super fond of this design, but I can't see any fix
> other than
> > refactoring the whole flags thing.
> 
> I suspected that... :-/ OK, then just add "(mksnapshot only)" to the
> descriptions.

Done.

[vogelheim, 2014-04-24 12:31:20]

Committed patchset #3 manually as r20941 (presubmit successful).