Adaptive padding prob distr tests #543
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
We need this for padding negotiation so that we can have later machine revisions supercede earlier ones.
This is a good code review start point, to get an overview of the interfaces and types used in circuit padding.
This helps us to determine if a middle node can pad to us or not.
These event callbacks allow circuit padding to decide when to attempt to launch and negotiate new padding machines, and when to tear old ones down.
These callbacks allow the padding state machines to react to various types of sent and received relay cells.
This implements all of the event handling, state machines, and padding decisions for circuit padding. I recommend reviewing this after you look at the call-in points into it from the rest of Tor.
* Tighten and justify cutoff for approximation in GPD sampler.
* Use expm1(-xi log U) to evaluate U^{-xi} - 1 without precision loss.
* Use log1p(x) to evaluate log(1 + x) precisely even for x near 0.
* Avoid 0 by rejection sampling, not by choosing to return 0 instead.
(This avoids the problematic 0 return from crypto_rand_double()
into the log() without introducing distribution bias).
Hope is this will make it easier to test on the live tor network. Does not need to be merged if we don't want to, but will come in handy for researchers.
Note to self/others: don't merge this.
Fix an appveyor fail.
Rename all the variables.
Rename all the variables.
Rename all the variables.
Factor end state cleanup code into its own function.
Fix issues with bin<->usec conversion.
Better tests for bin conversion.
Padding support is for version >= 1.
Clean up limit checks; also avoid overflow.
Add helper macro for infinity bin.
Use helper macro for infinity bin.
Make RTT calc slightly better.
Use range_usec instead of range_sec.
Fix negotiation failure due to not checking return values.
Invert transition array.
Invert transition array.
Invert transition array.
Add the code as is with no modifications as of Riastradh's commit b0f98cdb1d54c2b6b8f36cb183bbbc341a520f16 in https://mumble.net/~campbell/git/tor-dists.git.
- Shadowing warnings - Unused ok variable warnings - Disable some printfs that appear when tests pass.
This test only checks the uniform distribution and the way it's used by the circuitpadding subsystem. It doesn't test the other distributions.
asn-d6
force-pushed
the
adaptive_padding-prob-distr-tests
branch
from
83f78f8
to
ef86fbd
Compare
They were very slow (11s on my laptop) and also had 0.1% chance of failing.
There are various places where we legitimately test for equality: - comparison with zero for tests that should return zero - comparison with +/-infinity for tests that should return it - testing for probability p = 1 when p is limited to [0,1] There are also various places where we legitimately convert between integer and floating-point types: - choosing a geometric sample from ceil(log(...)) - choosing an integer bin from a continuous sample This patch: 1. Avoids unnecessary equality like p != 0 when p > 0 would work. 2. Uses silly comparisons like p >= 1 where p == 1 is meant. 3. Marks the silly cases with a comment `XXX -Wfloat-equal'. 4. Inserts explicit casts for intended conversion between integer and floating-point types. This patch does not address -Wbad-function-casts because it needlessly clutters up the code: when I write (size_t)floor((x - lo)/w), it's clear that I want the integer as an integer type, and clear _which_ integer I want.
Whitespace-only change: view with `git diff -b' to confirm.
The remainder are URLs that are too long to fit in 80 columns plus a comment marker.
1. For p near 0 or 1/2, there's nothing wrong with the translated cdf,
so enable these tests.
(For small x, i.e. p near 1/2, there would be a problem with
translation if we used something halfway between the cdf and the sf
that gives \int_0^x p instead of the cdf \int_{-\infty}^x p or the
sf \int_x^{+\infty} p. But we're not testing that.)
2. Fix the tests where either logit or logithalf is bad, so that we
test both if the input is the same and only the preciser one if
translation by 1/2 rounds one or the other.
3. Test logistic(-x) = 1 - p even if p == 1, as long as x is infinite.
I have a hard time following boolean parameters like that!
(only in tests, not for user input)
No reason to have an XXX over every -Wfloat-equal warning.
asn-d6
force-pushed
the
adaptive_padding-prob-distr-tests
branch
from
ef86fbd
to
cbe0d0e
Compare
Pull Request Test Coverage Report for Build 3103
💛 - Coveralls |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.