Timing regression?

I just ran the external testsuite for current master on my development
system, and I'm seeing some quite increased execution times:

    [ 50%] tests.short ... failed (+3.4%)
    [ 75%] tests.medium ... failed (+1.5%)
    [ 28%] tests.m57-short ... failed (+4.6%)
    [ 71%] tests.ipv6 ... failed (+39.8%)
    [ 85%] tests.m57-long ... failed (+9.8%)

The short tests are prone to fluctuate in timing, but the increases
for ipv6 and m57-long are sticking out. Any ideas what could be
causing this?


Did you ever happen to figure out what was going on with this?


No, but I also didn't look further. Could it be the new file
identifications (i.e., the regexps)?


That was my thought too. I'll have to look into DFA state creations to see if we've walked into that problem again.


Turns out it's the entropy analysis. The change below makes execution
times go down again.

That seems ok then? It's an optional policy script that the regression
testing pulls in, but users won't see a difference unless they decide
to load the script -- in which case it's reasonable to expect a
performance impact due to doing additional analysis.


--------- cut -------------------------------------------------------

diff --git a/scripts/policy/frameworks/files/entropy-test-all-files.bro b/scripts/policy/frameworks/files/entropy-test-all-files.bro
index fd02b9e..f66e2e9 100644
--- a/scripts/policy/frameworks/files/entropy-test-all-files.bro
+++ b/scripts/policy/frameworks/files/entropy-test-all-files.bro
@@ -11,10 +11,10 @@ export {

event file_new(f: fa_file)
- Files::add_analyzer(f, Files::ANALYZER_ENTROPY);
+ # Files::add_analyzer(f, Files::ANALYZER_ENTROPY);