Deep Memory Profiler: skip non-existing or empty files.

Deep Memory Profiler: skip non-existing or empty files.

Sometimes the dump series is not complete, or contains files with 0 bytes.

Rather than bailing out entirely, just skip those files.

This can happen for instance when using Telemetry-driven tests:
- each "page" on a "page set" is a different test, and will
contain its own set of "heap dumps"
- however, the "page set" itself is long running, so whilst
telemetry fetches heap dumps, chrome is still running and
generating dumps that aren't interesting.
- before starting the next page, telemetry removes all
those spurious dumps
- the final set of dump files contains "holes", but that's fine.

BUG=

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=202859

[bulach, 2013-05-24 15:47:23]

hey dai, this patch makes dmprof a bit more relaxed with the dump files, please
let me know if that's ok...

[Dai Mikurube (NOT FULLTIME), 2013-05-27 05:05:34]

Hmm, did you see a case that "middle" files are lost or empty?  I knew that the
last ".heap" file may be lost or empty while the ".buckets" file exists.  But, I
was not sure that it happens on non-last files...

[bulach, 2013-05-28 11:23:27]

on some telemetry-based tests, those "profilers" there fetch the set of files
per page, then clear the device directory.

if the measurement actually runs in the same browser across multiple pages,
there will be "holes"...

these missed files are not important because the measurement wasn't actually
running, i.e., browser and renderers are in a state that is not being measured.

so what happens is roughly:

for each page:
1) telemetry starts a "measurement"
2) actions are taken
3) telemetry finishes this "measurement"
4) telemetry profiler fetches all dump files, then clear the device directory..
the host contains a directory per page here..


So later on, when running dmp, sometimes it's useful to run it for all browser
dumps across all pages, even if we missed a few dumps when switching pages.

please let me know if that's fine.

[Dai Mikurube (NOT FULLTIME), 2013-05-29 04:43:05]

Makes sense.  LGTM.

Could you add some description of such a case in the commit log or the script
comment?

[bulach, 2013-05-29 10:15:14]

thanks dai! I edited the commit message, please let me know if it's any clear. 
feel free to edit it directly if you prefer!

[Dai Mikurube (NOT FULLTIME), 2013-05-29 10:28:02]

Looks good.  Please land it. :)

[commit-bot: I haz the power, 2013-05-29 10:45:03]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/bulach@chromium.org/15861007/1

[bulach, 2013-05-29 10:45:04]

thanks! :)

[commit-bot: I haz the power, 2013-05-29 13:52:07]

Change committed as 202859