support -aggregate for combining results from multiple processes

[derekbruening]

From bruen...@google.com on October 06, 2011 13:16:07

on linux I have an -aggregate option, used after running a tree of
processes that combines them all into one error results file. after
finishing issue #446 we should add that support. this would then eliminate
the need for the drmemory_analyze.py script.

Original issue: http://code.google.com/p/drmemory/issues/detail?id=614

[derekbruening]

From timurrrr@google.com on October 07, 2011 03:07:37

Aggregate = have just 1 results file instead of a bunch of them?

One of the reasons we use _analyze scripts for the other tools is that they can print per-test output.

That is, each test has its own BROWSER_WRAPPER PID. We group logfiles by the B_W PID and at the end of the run we may print out the reports grouped this way (= by a test case).
We'll need the script anyways at least for this grouping for that I think.

[derekbruening]

From bruen...@google.com on October 07, 2011 09:07:48

-aggregate as implemented on Linux takes in a list of logdirs and combines their results.txt files, resulting in a single file (this is not a trivial concat b/c the unique vs total is done properly, collapsing duplicates across the logdirs). it is up to the user to specify the logdirs. if I follow your description, you're saying you have many processes split into groups where each group corresponds to one test, so there's some logic to figure out which ones. even without that, somebody has to invoke the -aggregate after the runs are done, so there needs to be some script, but all the logic of reading and parsing results files should be in drmem code and doesn't need to be in an outside script.

[derekbruening]

From bruen...@google.com on March 09, 2012 08:41:25

Blocking: 827

[derekbruening]

From bruen...@google.com on March 15, 2012 08:12:58

xref issue #446 comment 5: while issue #446 perhaps can't be made cross-platform, we should be able to do everything but the symbolization cross-platform (i.e., combining and de-duping and suppression matching)

[derekbruening]

From bruen...@google.com on August 03, 2012 07:38:08

Blocking: drmemory:827

[derekbruening]

From bruen...@google.com on February 17, 2013 09:27:55

Labels: GoodContrib

[derekbruening]

From bruen...@google.com on April 21, 2014 10:49:49

Re-naming to reflect that we no longer have -aggregate on Linux due to switching to online symbols.

A first step here toward an automated identification of a process group is to add child pids to the log files: basically, re-enable the EXEC and other keyword log statements used for the old postprocess.pl model.

Summary: support -aggregate for combining results from multiple processes (was: support -aggregate on Windows)

[derekbruening]

Xref the script used for the Chromium waterfall: https://chromium.googlesource.com/chromium/src.git/+/master/tools/valgrind/drmemory_analyze.py