tor-commits November 2015

tor-commits@lists.torproject.org

20 participants
856 discussions

[webwml/staging] Update amazon button
by sebastian＠torproject.org 13 Nov '15

13 Nov '15

commit 172fab932427ebd028e83eca6e999073c172b25e Author: Sebastian Hahn <sebastian(a)torproject.org> Date: Fri Nov 13 04:07:02 2015 +0100 Update amazon button --- donate/en/donate-amazon.wml | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/donate/en/donate-amazon.wml b/donate/en/donate-amazon.wml index bcb96f4..0d4799f 100644 --- a/donate/en/donate-amazon.wml +++ b/donate/en/donate-amazon.wml @@ -17,7 +17,7 @@ src="https://static-na.payments-amazon.com/OffAmazonPayments/us/js/Widgets.js">< data-ap-widget-theme="ap-light" data-ap-widget-amount-presets="5,25,50,100" data-ap-widget-default-amount="1" - data-ap-signature="TccqAE9jXn74b3xQgYc5rFG9Y9N3Rbq3ihSQQfF0plg%3D" + data-ap-signature="%2B6UiARhqtXOPGLHP%2F6DrD3zH68ZHxmA9tBLjfjJ3jYU%3D" data-ap-seller-id="A3KD3XX49657E7" data-ap-access-key="AKIAISELZYVMUQ7F5M7Q" @@ -26,9 +26,9 @@ data-ap-lwa-client-id="amzn1.application-oa2-client.b0b3efe8eb6841ed800ff2cc45b1 data-ap-currency-code="USD" data-ap-amount="0" data-ap-note="Thank you for your donation to The Tor Project. The -Tor Project is a 501c3 organization, and our federal tax ID number is 20 -8096820. Your gift is tax deductible in the United States as allowed by -law. The Tor Project respects the confidentiality of its supporters, +Tor Project is a 501c3 organization, and our federal tax ID number is +20-8096820. Your gift is tax deductible in the United States as allowed +by law. The Tor Project respects the confidentiality of its supporters, and we do not lend, rent, or sell our lists of donors at any time. Your donation enables Tor developers to focus on making Tor better, including scalability, user support and better documentation, usability and @@ -37,7 +37,7 @@ like to keep in touch or get more involved, please visit us at www.torproject.org. Our mission is to advance human rights and freedoms by creating and deploying free and open anonymity and privacy technologies, supporting their unrestricted availability and use, and -furthering their scientific and popular understanding." +furthering their scientific and popular understanding. " data-ap-shipping-address-required="false" data-ap-payment-action="AuthorizeAndCapture" >

1 0

[translation/tails-greeter_completed] Update translations for tails-greeter_completed
by translation＠torproject.org 12 Nov '15

12 Nov '15

commit bf9573d989fd08fd12aa9beae46a4a0421deb25f Author: Translation commit bot <translation(a)torproject.org> Date: Thu Nov 12 22:45:31 2015 +0000 Update translations for tails-greeter_completed --- el/el.po | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) diff --git a/el/el.po b/el/el.po index ef5f2b0..7ca8865 100644 --- a/el/el.po +++ b/el/el.po @@ -8,14 +8,15 @@ # isv31 <36b043a6(a)anon.leemail.me>, 2014 # kotkotkot <kotakota(a)gmail.com>, 2012 # metamec, 2015 +# Natalia <hli8ia(a)gmail.com>, 2015 # Wasilis Mandratzis <inactive+Wasilis(a)transifex.com>, 2013 msgid "" msgstr "" "Project-Id-Version: The Tor Project\n" "Report-Msgid-Bugs-To: \n" -"POT-Creation-Date: 2015-08-05 18:46+0200\n" -"PO-Revision-Date: 2015-08-23 18:55+0000\n" -"Last-Translator: metamec\n" +"POT-Creation-Date: 2015-11-02 21:29+0100\n" +"PO-Revision-Date: 2015-11-12 22:33+0000\n" +"Last-Translator: Natalia <hli8ia(a)gmail.com>\n" "Language-Team: Greek (http://www.transifex.com/otf/torproject/language/el/)\n" "MIME-Version: 1.0\n" "Content-Type: text/plain; charset=UTF-8\n" @@ -163,6 +164,10 @@ msgid "" "need to configure bridge, firewall, or proxy settings." msgstr "Η σύνδεση διαδικτύου αυτού του υπολογιστή λογοκρίνεται, φιλτράρεται ή περνάει από proxy. Πρέπει να ρυθμίσετε τις επιλογές για τις γέφυρες, το firewall ή το proxy. " +#: ../glade/optionswindow.glade.h:23 +msgid "Disable all networking" +msgstr "Απενεργοποιήστε όλα τα δίκτυα" + #: ../glade/langpanel.glade.h:1 msgid " " msgstr " "

1 0

[translation/tails-greeter] Update translations for tails-greeter
by translation＠torproject.org 12 Nov '15

12 Nov '15

commit 1378b55a1d272aa04c53bc99e4c9432b5880cc10 Author: Translation commit bot <translation(a)torproject.org> Date: Thu Nov 12 22:45:25 2015 +0000 Update translations for tails-greeter --- el/el.po | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/el/el.po b/el/el.po index ed81f2a..7ca8865 100644 --- a/el/el.po +++ b/el/el.po @@ -8,14 +8,15 @@ # isv31 <36b043a6(a)anon.leemail.me>, 2014 # kotkotkot <kotakota(a)gmail.com>, 2012 # metamec, 2015 +# Natalia <hli8ia(a)gmail.com>, 2015 # Wasilis Mandratzis <inactive+Wasilis(a)transifex.com>, 2013 msgid "" msgstr "" "Project-Id-Version: The Tor Project\n" "Report-Msgid-Bugs-To: \n" -"POT-Creation-Date: 2015-10-26 17:51+0100\n" -"PO-Revision-Date: 2015-10-27 10:10+0000\n" -"Last-Translator: carolyn <carolyn(a)anhalt.org>\n" +"POT-Creation-Date: 2015-11-02 21:29+0100\n" +"PO-Revision-Date: 2015-11-12 22:33+0000\n" +"Last-Translator: Natalia <hli8ia(a)gmail.com>\n" "Language-Team: Greek (http://www.transifex.com/otf/torproject/language/el/)\n" "MIME-Version: 1.0\n" "Content-Type: text/plain; charset=UTF-8\n" @@ -165,7 +166,7 @@ msgstr "Η σύνδεση διαδικτύου αυτού του υπολογι #: ../glade/optionswindow.glade.h:23 msgid "Disable all networking" -msgstr "" +msgstr "Απενεργοποιήστε όλα τα δίκτυα" #: ../glade/langpanel.glade.h:1 msgid " "

1 0

[tor/master] Experimentally decouple the main body of circuit_mark_for_close
by nickm＠torproject.org 12 Nov '15

12 Nov '15

commit 8b4e5b7ee902fb7fa07767410a18433d752c7aef Author: Nick Mathewson <nickm(a)torproject.org> Date: Fri Oct 2 17:55:25 2015 +0200 Experimentally decouple the main body of circuit_mark_for_close --- changes/decouple_circuit_mark | 6 ++++ src/or/circuitlist.c | 77 ++++++++++++++++++++++++++++++----------- src/or/or.h | 8 +++++ 3 files changed, 71 insertions(+), 20 deletions(-) diff --git a/changes/decouple_circuit_mark b/changes/decouple_circuit_mark new file mode 100644 index 0000000..4b7ed77 --- /dev/null +++ b/changes/decouple_circuit_mark @@ -0,0 +1,6 @@ + o Code simplification and refactoring: + - Extract the more complicated parts of circuit_mark_for_close into + a new function run periodically before connections are freed. + This change removes more than half of the functions currently + in the "blob". + Closes ticket #17218. diff --git a/src/or/circuitlist.c b/src/or/circuitlist.c index 716024d..324f9f3 100644 --- a/src/or/circuitlist.c +++ b/src/or/circuitlist.c @@ -44,11 +44,16 @@ static smartlist_t *global_circuitlist = NULL; /** A list of all the circuits in CIRCUIT_STATE_CHAN_WAIT. */ static smartlist_t *circuits_pending_chans = NULL; +/** A list of all the circuits that have been marked with circuit_mark_for_close + * and which are waiting for circuit_about_to_free(). */ +static smartlist_t *circuits_pending_close = NULL; + static void circuit_free_cpath_node(crypt_path_t *victim); static void cpath_ref_decref(crypt_path_reference_t *cpath_ref); //static void circuit_set_rend_token(or_circuit_t *circ, int is_rend_circ, // const uint8_t *token); static void circuit_clear_rend_token(or_circuit_t *circ); +static void circuit_about_to_free(circuit_t *circ); /********* END VARIABLES ************/ @@ -451,16 +456,27 @@ circuit_count_pending_on_channel(channel_t *chan) void circuit_close_all_marked(void) { + if (circuits_pending_close == NULL) + return; + smartlist_t *lst = circuit_get_global_list(); - SMARTLIST_FOREACH_BEGIN(lst, circuit_t *, circ) { - /* Fix up index if SMARTLIST_DEL_CURRENT just moved this one. */ - circ->global_circuitlist_idx = circ_sl_idx; - if (circ->marked_for_close) { - circ->global_circuitlist_idx = -1; - circuit_free(circ); - SMARTLIST_DEL_CURRENT(lst, circ); + SMARTLIST_FOREACH_BEGIN(circuits_pending_close, circuit_t *, circ) { + tor_assert(circ->marked_for_close); + + /* Remove it from the circuit list. */ + int idx = circ->global_circuitlist_idx; + smartlist_del(lst, idx); + if (idx < smartlist_len(lst)) { + circuit_t *replacement = smartlist_get(lst, idx); + replacement->global_circuitlist_idx = idx; } + circ->global_circuitlist_idx = -1; + + circuit_about_to_free(circ); + circuit_free(circ); } SMARTLIST_FOREACH_END(circ); + + smartlist_clear(circuits_pending_close); } /** Return the head of the global linked list of circuits. */ @@ -1703,6 +1719,39 @@ circuit_mark_for_close_, (circuit_t *circ, int reason, int line, reason = END_CIRC_REASON_NONE; } + circ->marked_for_close = line; + circ->marked_for_close_file = file; + circ->marked_for_close_reason = reason; + circ->marked_for_close_orig_reason = orig_reason; + + if (!CIRCUIT_IS_ORIGIN(circ)) { + or_circuit_t *or_circ = TO_OR_CIRCUIT(circ); + if (or_circ->rend_splice) { + if (!or_circ->rend_splice->base_.marked_for_close) { + /* do this after marking this circuit, to avoid infinite recursion. */ + circuit_mark_for_close(TO_CIRCUIT(or_circ->rend_splice), reason); + } + or_circ->rend_splice = NULL; + } + } + + if (circuits_pending_close == NULL) + circuits_pending_close = smartlist_new(); + + smartlist_add(circuits_pending_close, circ); +} + +/** Called immediately before freeing a marked circuit circ. + * Disconnects the circuit from other data structures, launches events + * as appropriate, and performs other housekeeping. + */ +static void +circuit_about_to_free(circuit_t *circ) +{ + + int reason = circ->marked_for_close_reason; + int orig_reason = circ->marked_for_close_orig_reason; + if (circ->state == CIRCUIT_STATE_ONIONSKIN_PENDING) { onion_pending_remove(TO_OR_CIRCUIT(circ)); } @@ -1726,6 +1775,7 @@ circuit_mark_for_close_, (circuit_t *circ, int reason, int line, (circ->state == CIRCUIT_STATE_OPEN)?CIRC_EVENT_CLOSED:CIRC_EVENT_FAILED, orig_reason); } + if (circ->purpose == CIRCUIT_PURPOSE_C_INTRODUCE_ACK_WAIT) { origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ); int timed_out = (reason == END_CIRC_REASON_TIMEOUT); @@ -1811,19 +1861,6 @@ circuit_mark_for_close_, (circuit_t *circ, int reason, int line, ocirc->p_streams = NULL; } - circ->marked_for_close = line; - circ->marked_for_close_file = file; - - if (!CIRCUIT_IS_ORIGIN(circ)) { - or_circuit_t *or_circ = TO_OR_CIRCUIT(circ); - if (or_circ->rend_splice) { - if (!or_circ->rend_splice->base_.marked_for_close) { - /* do this after marking this circuit, to avoid infinite recursion. */ - circuit_mark_for_close(TO_CIRCUIT(or_circ->rend_splice), reason); - } - or_circ->rend_splice = NULL; - } - } } /** Given a marked circuit circ, aggressively free its cell queues to diff --git a/src/or/or.h b/src/or/or.h index a80cd55..746e462 100644 --- a/src/or/or.h +++ b/src/or/or.h @@ -2892,6 +2892,14 @@ typedef struct circuit_t { * where this circuit was marked.) */ const char *marked_for_close_file; /**< For debugging: in which file was this * circuit marked for close? */ + /** For what reason (See END_CIRC_REASON...) is this circuit being closed? + * This field is set in circuit_mark_for_close and used later in + * circuit_about_to_free. */ + uint16_t marked_for_close_reason; + /** As marked_for_close_reason, but reflects the underlying reason for + * closing this circuit. + */ + uint16_t marked_for_close_orig_reason; /** Unique ID for measuring tunneled network status requests. */ uint64_t dirreq_id;

1 0

[tor/master] Merge branch 'decouple_circuit_mark_squashed'
by nickm＠torproject.org 12 Nov '15

12 Nov '15

commit f7ccc9b975ae3e6531767c1a56cd85c906292aff Merge: b370052 8b4e5b7 Author: Nick Mathewson <nickm(a)torproject.org> Date: Thu Nov 12 14:20:24 2015 -0500 Merge branch 'decouple_circuit_mark_squashed' changes/decouple_circuit_mark | 6 ++++ src/or/circuitlist.c | 77 ++++++++++++++++++++++++++++++----------- src/or/or.h | 8 +++++ 3 files changed, 71 insertions(+), 20 deletions(-)

1 0

[tor/master] + and / usage clarification - Fixes #13158
by nickm＠torproject.org 12 Nov '15

12 Nov '15

commit b370052ae418284a6f75a15708d2d0685ec765f4 Author: Joan Queralt <jqueraltmo(a)gmail.com> Date: Mon Nov 2 01:36:19 2015 +0100 + and / usage clarification - Fixes #13158 --- doc/tor.1.txt | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/doc/tor.1.txt b/doc/tor.1.txt index 916433b..04826b3 100644 --- a/doc/tor.1.txt +++ b/doc/tor.1.txt @@ -128,13 +128,16 @@ complicated for options that are allowed to occur more than once: if you specify four SOCKSPorts in your configuration file, and one more SOCKSPort on the command line, the option on the command line will replace __all__ of the SOCKSPorts in the configuration file. If this isn't what you want, prefix -the option name with a plus sign, and it will be appended to the previous set -of options instead. +the option name with a plus sign (+), and it will be appended to the previous +set of options instead. For example, setting SOCKSPort 9100 will use only +port 9100, but setting +SOCKSPort 9100 will use ports 9100 and 9050 (because +this is the default). Alternatively, you might want to remove every instance of an option in the configuration file, and not replace it at all: you might want to say on the command line that you want no SOCKSPorts at all. To do that, prefix the -option name with a forward slash. +option name with a forward slash (/). You can use the plus sign (+) and the +forward slash (/) in the configuration file and on the command line. GENERAL OPTIONS --------------- @@ -2557,7 +2560,7 @@ __DataDirectory__**/control_auth_cookie**:: overridden by the CookieAuthFile config option. Regenerated on startup. See control-spec.txt in https://spec.torproject.org/[torspec] for details. Only used when cookie authentication is enabled. - + __DataDirectory__**/lock**:: This file is used to prevent two Tor instances from using same data directory. If access to this file is locked, data directory is already

1 0

[tor/master] Handle edge cases in laplace functions
by nickm＠torproject.org 12 Nov '15

12 Nov '15

commit 6d54bdbdcf076167c1b73bfb5bef9fd1c3921796 Author: teor <teor2345(a)gmail.com> Date: Thu Dec 25 20:30:18 2014 +1100 Handle edge cases in laplace functions Avoid division by zero. Avoid taking the log of zero. Silence clang type conversion warnings using round and trunc. The existing values returned by the laplace functions do not change. Add tests for laplace edge cases. These changes pass the existing unit tests without modification. Related to HS stats in #13192. --- changes/laplace-edge-cases | 7 ++ src/common/util.c | 33 ++++++-- src/test/test_util.c | 201 ++++++++++++++++++++++++++++++++++++++++++-- 3 files changed, 226 insertions(+), 15 deletions(-) diff --git a/changes/laplace-edge-cases b/changes/laplace-edge-cases new file mode 100644 index 0000000..f8a36e1 --- /dev/null +++ b/changes/laplace-edge-cases @@ -0,0 +1,7 @@ + o Minor bugfixes: + - Handle edge cases in the laplace functions: + * avoid division by zero + * avoid taking the log of zero + * silence clang type conversion warnings using round and trunc + - Add tests for laplace edge cases. + Related to HS stats in #13192. diff --git a/src/common/util.c b/src/common/util.c index f8d1b7b..52b3e04 100644 --- a/src/common/util.c +++ b/src/common/util.c @@ -536,13 +536,18 @@ int64_t sample_laplace_distribution(double mu, double b, double p) { double result; - tor_assert(p >= 0.0 && p < 1.0); + /* This is the "inverse cumulative distribution function" from: * http://en.wikipedia.org/wiki/Laplace_distribution */ + if (p == 0.0) { + /* Avoid taking log(0.0) == -INFINITY, as some processors or compiler + * options can cause the program to trap. */ + return INT64_MIN; + } + result = mu - b * (p > 0.5 ? 1.0 : -1.0) * tor_mathlog(1.0 - 2.0 * fabs(p - 0.5)); - if (result >= INT64_MAX) return INT64_MAX; else if (result <= INT64_MIN) @@ -551,18 +556,28 @@ sample_laplace_distribution(double mu, double b, double p) return (int64_t) result; } -/** Add random noise between INT64_MIN and INT64_MAX coming from a - * Laplace distribution with mu = 0 and b = delta_f/epsilon - * to signal based on the provided random value in - * [0.0, 1.0[. */ +/** Add random noise between INT64_MIN and INT64_MAX coming from a Laplace + * distribution with mu = 0 and b = delta_f/epsilon to + * signal based on the provided random value in [0.0, 1.0[. + * The epislon value must be between ]0.0, 1.0]. delta_f must be greater + * than 0. */ int64_t add_laplace_noise(int64_t signal, double random, double delta_f, double epsilon) { - int64_t noise = sample_laplace_distribution( - 0.0, /* just add noise, no further signal */ - delta_f / epsilon, random); + int64_t noise; + + /* epsilon MUST be between ]0.0, 1.0] */ + tor_assert(epsilon > 0.0 && epsilon <= 1.0); + /* delta_f MUST be greater than 0. */ + tor_assert(delta_f > 0.0); + + /* Just add noise, no further signal */ + noise = sample_laplace_distribution(0.0, + delta_f / epsilon, + random); + /* Clip (signal + noise) to [INT64_MIN, INT64_MAX] */ if (noise > 0 && INT64_MAX - noise < signal) return INT64_MAX; else if (noise < 0 && INT64_MIN - noise > signal) diff --git a/src/test/test_util.c b/src/test/test_util.c index 6a4c3ec..940f9bd 100644 --- a/src/test/test_util.c +++ b/src/test/test_util.c @@ -4100,12 +4100,201 @@ test_util_laplace(void *arg) */ tt_i64_op(INT64_MIN + 20, ==, add_laplace_noise(20, 0.0, delta_f, epsilon)); - tt_i64_op(-60, ==, add_laplace_noise(20, 0.1, delta_f, epsilon)); - tt_i64_op(-14, ==, add_laplace_noise(20, 0.25, delta_f, epsilon)); - tt_i64_op(20, ==, add_laplace_noise(20, 0.5, delta_f, epsilon)); - tt_i64_op(54, ==, add_laplace_noise(20, 0.75, delta_f, epsilon)); - tt_i64_op(100, ==, add_laplace_noise(20, 0.9, delta_f, epsilon)); - tt_i64_op(215, ==, add_laplace_noise(20, 0.99, delta_f, epsilon)); + + tt_assert(-60 == add_laplace_noise(20, 0.1, delta_f, epsilon)); + tt_assert(-14 == add_laplace_noise(20, 0.25, delta_f, epsilon)); + tt_assert(20 == add_laplace_noise(20, 0.5, delta_f, epsilon)); + tt_assert(54 == add_laplace_noise(20, 0.75, delta_f, epsilon)); + tt_assert(100 == add_laplace_noise(20, 0.9, delta_f, epsilon)); + tt_assert(215 == add_laplace_noise(20, 0.99, delta_f, epsilon)); + + /* Test extreme values of signal with maximally negative values of noise + * 1.0000000000000002 is the smallest number > 1 + * 0.0000000000000002 is the double epsilon (error when calculating near 1) + * this is approximately 1/(2^52) + * per https://en.wikipedia.org/wiki/Double_precision + * (let's not descend into the world of subnormals) + * >>> laplace.ppf([0, 0.0000000000000002], loc = 0, scale = 1) + * array([ -inf, -35.45506713]) + */ + const double noscale_df = 1.0, noscale_eps = 1.0; + + tt_assert(INT64_MIN == + add_laplace_noise(0, 0.0, noscale_df, noscale_eps)); + + /* is it clipped to INT64_MIN? */ + tt_assert(INT64_MIN == + add_laplace_noise(-1, 0.0, noscale_df, noscale_eps)); + tt_assert(INT64_MIN == + add_laplace_noise(INT64_MIN, 0.0, + noscale_df, noscale_eps)); + /* ... even when scaled? */ + tt_assert(INT64_MIN == + add_laplace_noise(0, 0.0, delta_f, epsilon)); + tt_assert(INT64_MIN == + add_laplace_noise(0, 0.0, + INT64_MAX, 1)); + tt_assert(INT64_MIN == + add_laplace_noise(INT64_MIN, 0.0, + INT64_MAX, 1)); + + /* does it play nice with INT64_MAX? */ + tt_assert((INT64_MIN + INT64_MAX) == + add_laplace_noise(INT64_MAX, 0.0, + noscale_df, noscale_eps)); + + /* do near-zero fractional values work? */ + const double min_dbl_error = 0.0000000000000002; + + tt_assert(-35 == + add_laplace_noise(0, min_dbl_error, + noscale_df, noscale_eps)); + tt_assert(INT64_MIN == + add_laplace_noise(INT64_MIN, min_dbl_error, + noscale_df, noscale_eps)); + tt_assert((-35 + INT64_MAX) == + add_laplace_noise(INT64_MAX, min_dbl_error, + noscale_df, noscale_eps)); + /* ... even when scaled? */ + tt_assert(INT64_MIN == + add_laplace_noise(0, min_dbl_error, + INT64_MIN, -35)); + tt_assert(INT64_MIN == + add_laplace_noise(0, min_dbl_error, + INT64_MIN, -34)); + tt_assert(INT64_MAX == + add_laplace_noise(0, min_dbl_error, + INT64_MIN, 1)); + tt_assert((INT64_MIN + INT64_MAX) == + add_laplace_noise(INT64_MIN, min_dbl_error, + INT64_MIN, 1)); + tt_assert(INT64_MAX == + add_laplace_noise(INT64_MAX, min_dbl_error, + INT64_MIN, 1)); + + tt_assert(INT64_MAX == + add_laplace_noise(0, min_dbl_error, + INT64_MAX, -35)); + tt_assert(INT64_MAX == + add_laplace_noise(0, min_dbl_error, + INT64_MAX, -34)); + tt_assert(INT64_MIN == + add_laplace_noise(0, min_dbl_error, + INT64_MAX, 1)); + tt_assert((INT64_MAX + INT64_MIN) == + add_laplace_noise(INT64_MAX, min_dbl_error, + INT64_MAX, 1)); + tt_assert(INT64_MIN == + add_laplace_noise(INT64_MIN, min_dbl_error, + INT64_MAX, 1)); + + /* does it play nice with INT64_MAX? */ + tt_assert((INT64_MAX - 35) == + add_laplace_noise(INT64_MAX, min_dbl_error, + noscale_df, noscale_eps)); + + /* Test extreme values of signal with maximally positive values of noise + * 1.0000000000000002 is the smallest number > 1 + * 0.9999999999999998 is the greatest number < 1 by calculation + * per https://en.wikipedia.org/wiki/Double_precision + * >>> laplace.ppf([1.0, 0.9999999999999998], loc = 0, scale = 1) + * array([inf, 35.35050621]) + * but the function rejects p == 1.0, so we just use max_dbl_lt_one + */ + const double max_dbl_lt_one = 0.9999999999999998; + + /* do near-one fractional values work? */ + tt_assert(35 == + add_laplace_noise(0, max_dbl_lt_one, noscale_df, noscale_eps)); + + /* is it clipped to INT64_MAX? */ + tt_assert(INT64_MAX == + add_laplace_noise(INT64_MAX - 35, max_dbl_lt_one, + noscale_df, noscale_eps)); + tt_assert(INT64_MAX == + add_laplace_noise(INT64_MAX - 34, max_dbl_lt_one, + noscale_df, noscale_eps)); + tt_assert(INT64_MAX == + add_laplace_noise(INT64_MAX, max_dbl_lt_one, + noscale_df, noscale_eps)); + /* ... even when scaled? */ + tt_assert(INT64_MAX == + add_laplace_noise(INT64_MAX, max_dbl_lt_one, + delta_f, epsilon)); + tt_assert(INT64_MAX == + add_laplace_noise(0, max_dbl_lt_one, + INT64_MAX, 35)); + tt_assert(INT64_MAX == + add_laplace_noise(0, max_dbl_lt_one, + INT64_MAX, 34)); + tt_assert((INT64_MIN + INT64_MAX) == + add_laplace_noise(INT64_MIN, max_dbl_lt_one, + INT64_MAX, 1)); + tt_assert(INT64_MAX == + add_laplace_noise(INT64_MAX, max_dbl_lt_one, + INT64_MAX, 1)); + tt_assert((INT64_MAX + INT64_MIN) == + add_laplace_noise(INT64_MAX, max_dbl_lt_one, + INT64_MIN, 1)); + tt_assert(INT64_MIN == + add_laplace_noise(INT64_MIN, max_dbl_lt_one, + INT64_MIN, 1)); + + /* does it play nice with INT64_MIN? */ + tt_assert((INT64_MIN + 35) == + add_laplace_noise(INT64_MIN, max_dbl_lt_one, + noscale_df, noscale_eps)); + + /* Test extreme values of b = delta_f / epsilon + * >>> laplace.ppf([0.5], loc = 0, scale = 1) + * array([0. ]) + */ + + /* Make sure edge cases don't depend on architecture, + * optimisation level, or other compiler flags. + * XXXX Are these edge cases important enough to make consistent? */ + + /* b = positive zero, p yields positive zero */ + tt_assert(0.0 == + add_laplace_noise(0.0, 0.5, 0.0, 1.0)) + /* b = negative zero, p yields positive zero */ + tt_assert(0.0 == + add_laplace_noise(0.0, 0.5, 0.0, -1.0)) + /* b = positive infinity, p yields positive zero, result is -NaN -> -Inf */ + tt_assert(INT64_MIN == + add_laplace_noise(0.0, 0.5, 1.0, 0.0)) + /* b = negative infinity, p yields positive zero, result is -NaN -> -Inf */ + tt_assert(INT64_MIN == + add_laplace_noise(0.0, 0.5, -1.0, 0.0)) + /* b = positive NaN (rounded to -Inf), p yields positive zero, + * result is -NaN -> -Inf */ + tt_assert(INT64_MIN == + add_laplace_noise(0.0, 0.5, -0.0, -0.0)) + /* b = negative NaN (rounded to -Inf), p yields positive zero, + * result is -NaN -> -Inf*/ + tt_assert(INT64_MIN == + add_laplace_noise(0.0, 0.5, -0.0, 0.0)) + + /* b = positive zero, p yields negative infinity, result is -NaN -> -Inf */ + tt_assert(INT64_MIN == + add_laplace_noise(0.0, 0.0, 0.0, 1.0)) + /* b = negative zero, p yields negative infinity, result is -NaN -> -Inf */ + tt_assert(INT64_MIN == + add_laplace_noise(0.0, 0.0, 0.0, -1.0)) + /* b = positive infinity, p yields negative infinity */ + tt_assert(INT64_MIN == + add_laplace_noise(0.0, 0.0, 1.0, 0.0)) + /* b = negative infinity, p yields negative infinity */ + tt_assert(INT64_MAX == + add_laplace_noise(0.0, 0.0, -1.0, 0.0)) + /* b = positive NaN (rounded to -Inf), p yields negative infinity, + * result is -NaN -> -Inf */ + tt_assert(INT64_MIN == + add_laplace_noise(0.0, 0.0, -0.0, -0.0)) + /* b = negative NaN (rounded to -Inf), p yields negative infinity, + * result is NaN -> Inf */ + tt_assert(INT64_MAX == + add_laplace_noise(0.0, 0.0, -0.0, 0.0)) done: ;

1 0

[tor/master] Handle edge cases in the round_*_to_next_multiple_of functions
by nickm＠torproject.org 12 Nov '15

12 Nov '15

commit 09cac24373e5a13cc527bf2f32132a9479d4ae1e Author: teor <teor2345(a)gmail.com> Date: Thu Dec 25 20:52:10 2014 +1100 Handle edge cases in the round_*_to_next_multiple_of functions Consistently check for overflow in round_*_to_next_multiple_of. Check all round_*_to_next_multiple_of functions with expected values. Check all round_*_to_next_multiple_of functions with maximal values. Related to HS stats in #13192. --- changes/laplace-edge-cases | 6 +++++- src/common/util.c | 12 +++++++++--- src/test/test_util.c | 27 +++++++++++++++++++++++++-- 3 files changed, 39 insertions(+), 6 deletions(-) diff --git a/changes/laplace-edge-cases b/changes/laplace-edge-cases index f8a36e1..266b119 100644 --- a/changes/laplace-edge-cases +++ b/changes/laplace-edge-cases @@ -3,5 +3,9 @@ * avoid division by zero * avoid taking the log of zero * silence clang type conversion warnings using round and trunc - - Add tests for laplace edge cases. + * consistently check for overflow in round_*_to_next_multiple_of + - Add tests for laplace edge cases: + * check add_laplace_noise with maximal values + * check round_*_to_next_multiple_of with additional values + * check round_*_to_next_multiple_of with maximal values Related to HS stats in #13192. diff --git a/src/common/util.c b/src/common/util.c index 52b3e04..3e680d2 100644 --- a/src/common/util.c +++ b/src/common/util.c @@ -491,7 +491,9 @@ round_to_power_of_2(uint64_t u64) unsigned round_to_next_multiple_of(unsigned number, unsigned divisor) { - number += divisor - 1; + tor_assert(divisor > 0); + if (UINT_MAX - divisor + 1 >= number) + number += divisor - 1; number -= number % divisor; return number; } @@ -501,7 +503,9 @@ round_to_next_multiple_of(unsigned number, unsigned divisor) uint32_t round_uint32_to_next_multiple_of(uint32_t number, uint32_t divisor) { - number += divisor - 1; + tor_assert(divisor > 0); + if (UINT32_MAX - divisor + 1 >= number) + number += divisor - 1; number -= number % divisor; return number; } @@ -511,7 +515,9 @@ round_uint32_to_next_multiple_of(uint32_t number, uint32_t divisor) uint64_t round_uint64_to_next_multiple_of(uint64_t number, uint64_t divisor) { - number += divisor - 1; + tor_assert(divisor > 0); + if (UINT64_MAX - divisor + 1 >= number) + number += divisor - 1; number -= number % divisor; return number; } diff --git a/src/test/test_util.c b/src/test/test_util.c index 940f9bd..74d7227 100644 --- a/src/test/test_util.c +++ b/src/test/test_util.c @@ -4054,8 +4054,11 @@ test_util_round_to_next_multiple_of(void *arg) tt_u64_op(round_uint64_to_next_multiple_of(99,7), ==, 105); tt_u64_op(round_uint64_to_next_multiple_of(99,9), ==, 99); - tt_i64_op(round_int64_to_next_multiple_of(0,1), ==, 0); - tt_i64_op(round_int64_to_next_multiple_of(0,7), ==, 0); + tt_assert(round_uint64_to_next_multiple_of(UINT64_MAX,2) == + UINT64_MAX-UINT64_MAX%2); + + tt_assert(round_int64_to_next_multiple_of(0,1) == 0); + tt_assert(round_int64_to_next_multiple_of(0,7) == 0); tt_i64_op(round_int64_to_next_multiple_of(99,1), ==, 99); tt_i64_op(round_int64_to_next_multiple_of(99,7), ==, 105); @@ -4068,6 +4071,26 @@ test_util_round_to_next_multiple_of(void *arg) tt_i64_op(round_int64_to_next_multiple_of(INT64_MIN,2), ==, INT64_MIN); tt_i64_op(round_int64_to_next_multiple_of(INT64_MAX,2), ==, INT64_MAX-INT64_MAX%2); + + tt_assert(round_uint32_to_next_multiple_of(0,1) == 0); + tt_assert(round_uint32_to_next_multiple_of(0,7) == 0); + + tt_assert(round_uint32_to_next_multiple_of(99,1) == 99); + tt_assert(round_uint32_to_next_multiple_of(99,7) == 105); + tt_assert(round_uint32_to_next_multiple_of(99,9) == 99); + + tt_assert(round_uint32_to_next_multiple_of(UINT32_MAX,2) == + UINT32_MAX-UINT32_MAX%2); + + tt_assert(round_to_next_multiple_of(0,1) == 0); + tt_assert(round_to_next_multiple_of(0,7) == 0); + + tt_assert(round_to_next_multiple_of(99,1) == 99); + tt_assert(round_to_next_multiple_of(99,7) == 105); + tt_assert(round_to_next_multiple_of(99,9) == 99); + + tt_assert(round_to_next_multiple_of(UINT_MAX,2) == + UINT_MAX-UINT_MAX%2); done: ; }

1 0

[tor/master] Test: change INT64_MAX to DBL_MAX in laplace test
by nickm＠torproject.org 12 Nov '15

12 Nov '15

commit dcf0f808c0bd9b20a0a7d4f3e52155a83cdce2a4 Author: David Goulet <dgoulet(a)ev0ke.net> Date: Thu Feb 12 11:36:35 2015 -0500 Test: change INT64_MAX to DBL_MAX in laplace test Signed-off-by: David Goulet <dgoulet(a)ev0ke.net> --- src/test/test_util.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/src/test/test_util.c b/src/test/test_util.c index 21737f6..2f05ce8 100644 --- a/src/test/test_util.c +++ b/src/test/test_util.c @@ -19,6 +19,7 @@ #endif #include <math.h> #include <ctype.h> +#include <values.h> /* XXXX this is a minimal wrapper to make the unit tests compile with the * changed tor_timegm interface. */ @@ -4156,10 +4157,10 @@ test_util_laplace(void *arg) add_laplace_noise(0, 0.0, delta_f, epsilon)); tt_assert(INT64_MIN == add_laplace_noise(0, 0.0, - INT64_MAX, 1)); + DBL_MAX, 1)); tt_assert(INT64_MIN == add_laplace_noise(INT64_MIN, 0.0, - INT64_MAX, 1)); + DBL_MAX, 1)); /* does it play nice with INT64_MAX? */ tt_assert((INT64_MIN + INT64_MAX) == @@ -4180,13 +4181,13 @@ test_util_laplace(void *arg) noscale_df, noscale_eps)); tt_assert(INT64_MIN == add_laplace_noise(0, min_dbl_error, - INT64_MAX, 1)); + DBL_MAX, 1)); tt_assert((INT64_MAX + INT64_MIN) == add_laplace_noise(INT64_MAX, min_dbl_error, - INT64_MAX, 1)); + DBL_MAX, 1)); tt_assert(INT64_MIN == add_laplace_noise(INT64_MIN, min_dbl_error, - INT64_MAX, 1)); + DBL_MAX, 1)); /* does it play nice with INT64_MAX? */ tt_assert((INT64_MAX - 35) == @@ -4223,10 +4224,10 @@ test_util_laplace(void *arg) delta_f, epsilon)); tt_assert((INT64_MIN + INT64_MAX) == add_laplace_noise(INT64_MIN, max_dbl_lt_one, - INT64_MAX, 1)); + DBL_MAX, 1)); tt_assert(INT64_MAX == add_laplace_noise(INT64_MAX, max_dbl_lt_one, - INT64_MAX, 1)); + DBL_MAX, 1)); /* does it play nice with INT64_MIN? */ tt_assert((INT64_MIN + 35) == add_laplace_noise(INT64_MIN, max_dbl_lt_one,

1 0

[tor/master] Tweak teor's and dgoulet's #13192 patches.
by nickm＠torproject.org 12 Nov '15

12 Nov '15

commit dad5eb7e1f3c92974c7bcaaea325a05233a27494 Author: Karsten Loesing <karsten.loesing(a)gmx.net> Date: Fri Feb 13 11:32:35 2015 +0100 Tweak teor's and dgoulet's #13192 patches. - Rewrite changes file. - Avoid float comparison with == and use <= instead. - Add teor's tor_llround(trunc(...)) back to silence clang warnings. - Replace tt_assert() with tt_i64_op() and friends. - Fix whitespace and a comment. --- changes/laplace-edge-cases | 19 +++++----- src/common/util.c | 10 +++--- src/test/test_util.c | 86 ++++++++++++++++++++++---------------------- 3 files changed, 56 insertions(+), 59 deletions(-) diff --git a/changes/laplace-edge-cases b/changes/laplace-edge-cases index 266b119..6c8c77b 100644 --- a/changes/laplace-edge-cases +++ b/changes/laplace-edge-cases @@ -1,11 +1,8 @@ - o Minor bugfixes: - - Handle edge cases in the laplace functions: - * avoid division by zero - * avoid taking the log of zero - * silence clang type conversion warnings using round and trunc - * consistently check for overflow in round_*_to_next_multiple_of - - Add tests for laplace edge cases: - * check add_laplace_noise with maximal values - * check round_*_to_next_multiple_of with additional values - * check round_*_to_next_multiple_of with maximal values - Related to HS stats in #13192. + o Code simplifications and unit tests: + - Handle edge cases in the laplace functions: avoid division by zero, + avoid taking the log of zero, and silence clang type conversion + warnings using round and trunc. Add unit tests for edge cases with + maximal values. + - Consistently check for overflow in round_*_to_next_multiple_of + functions, and add unit tests with additional and maximal values. + diff --git a/src/common/util.c b/src/common/util.c index 3e680d2..d8da8b1 100644 --- a/src/common/util.c +++ b/src/common/util.c @@ -546,26 +546,26 @@ sample_laplace_distribution(double mu, double b, double p) /* This is the "inverse cumulative distribution function" from: * http://en.wikipedia.org/wiki/Laplace_distribution */ - if (p == 0.0) { + if (p <= 0.0) { /* Avoid taking log(0.0) == -INFINITY, as some processors or compiler * options can cause the program to trap. */ return INT64_MIN; } - result = mu - b * (p > 0.5 ? 1.0 : -1.0) - * tor_mathlog(1.0 - 2.0 * fabs(p - 0.5)); + result = mu - b * (p > 0.5 ? 1.0 : -1.0) + * tor_mathlog(1.0 - 2.0 * fabs(p - 0.5)); if (result >= INT64_MAX) return INT64_MAX; else if (result <= INT64_MIN) return INT64_MIN; else - return (int64_t) result; + return tor_llround(trunc(result)); } /** Add random noise between INT64_MIN and INT64_MAX coming from a Laplace * distribution with mu = 0 and b = delta_f/epsilon to * signal based on the provided random value in [0.0, 1.0[. - * The epislon value must be between ]0.0, 1.0]. delta_f must be greater + * The epsilon value must be between ]0.0, 1.0]. delta_f must be greater * than 0. */ int64_t add_laplace_noise(int64_t signal, double random, double delta_f, diff --git a/src/test/test_util.c b/src/test/test_util.c index 2f05ce8..3310316 100644 --- a/src/test/test_util.c +++ b/src/test/test_util.c @@ -4055,11 +4055,11 @@ test_util_round_to_next_multiple_of(void *arg) tt_u64_op(round_uint64_to_next_multiple_of(99,7), ==, 105); tt_u64_op(round_uint64_to_next_multiple_of(99,9), ==, 99); - tt_assert(round_uint64_to_next_multiple_of(UINT64_MAX,2) == + tt_u64_op(round_uint64_to_next_multiple_of(UINT64_MAX,2), ==, UINT64_MAX-UINT64_MAX%2); - tt_assert(round_int64_to_next_multiple_of(0,1) == 0); - tt_assert(round_int64_to_next_multiple_of(0,7) == 0); + tt_i64_op(round_int64_to_next_multiple_of(0,1), ==, 0); + tt_i64_op(round_int64_to_next_multiple_of(0,7), ==, 0); tt_i64_op(round_int64_to_next_multiple_of(99,1), ==, 99); tt_i64_op(round_int64_to_next_multiple_of(99,7), ==, 105); @@ -4073,24 +4073,24 @@ test_util_round_to_next_multiple_of(void *arg) tt_i64_op(round_int64_to_next_multiple_of(INT64_MAX,2), ==, INT64_MAX-INT64_MAX%2); - tt_assert(round_uint32_to_next_multiple_of(0,1) == 0); - tt_assert(round_uint32_to_next_multiple_of(0,7) == 0); + tt_int_op(round_uint32_to_next_multiple_of(0,1), ==, 0); + tt_int_op(round_uint32_to_next_multiple_of(0,7), ==, 0); - tt_assert(round_uint32_to_next_multiple_of(99,1) == 99); - tt_assert(round_uint32_to_next_multiple_of(99,7) == 105); - tt_assert(round_uint32_to_next_multiple_of(99,9) == 99); + tt_int_op(round_uint32_to_next_multiple_of(99,1), ==, 99); + tt_int_op(round_uint32_to_next_multiple_of(99,7), ==, 105); + tt_int_op(round_uint32_to_next_multiple_of(99,9), ==, 99); - tt_assert(round_uint32_to_next_multiple_of(UINT32_MAX,2) == + tt_int_op(round_uint32_to_next_multiple_of(UINT32_MAX,2), ==, UINT32_MAX-UINT32_MAX%2); - tt_assert(round_to_next_multiple_of(0,1) == 0); - tt_assert(round_to_next_multiple_of(0,7) == 0); + tt_uint_op(round_to_next_multiple_of(0,1), ==, 0); + tt_uint_op(round_to_next_multiple_of(0,7), ==, 0); - tt_assert(round_to_next_multiple_of(99,1) == 99); - tt_assert(round_to_next_multiple_of(99,7) == 105); - tt_assert(round_to_next_multiple_of(99,9) == 99); + tt_uint_op(round_to_next_multiple_of(99,1), ==, 99); + tt_uint_op(round_to_next_multiple_of(99,7), ==, 105); + tt_uint_op(round_to_next_multiple_of(99,9), ==, 99); - tt_assert(round_to_next_multiple_of(UINT_MAX,2) == + tt_uint_op(round_to_next_multiple_of(UINT_MAX,2), ==, UINT_MAX-UINT_MAX%2); done: ; @@ -4125,12 +4125,12 @@ test_util_laplace(void *arg) tt_i64_op(INT64_MIN + 20, ==, add_laplace_noise(20, 0.0, delta_f, epsilon)); - tt_assert(-60 == add_laplace_noise(20, 0.1, delta_f, epsilon)); - tt_assert(-14 == add_laplace_noise(20, 0.25, delta_f, epsilon)); - tt_assert(20 == add_laplace_noise(20, 0.5, delta_f, epsilon)); - tt_assert(54 == add_laplace_noise(20, 0.75, delta_f, epsilon)); - tt_assert(100 == add_laplace_noise(20, 0.9, delta_f, epsilon)); - tt_assert(215 == add_laplace_noise(20, 0.99, delta_f, epsilon)); + tt_i64_op(-60, ==, add_laplace_noise(20, 0.1, delta_f, epsilon)); + tt_i64_op(-14, ==, add_laplace_noise(20, 0.25, delta_f, epsilon)); + tt_i64_op(20, ==, add_laplace_noise(20, 0.5, delta_f, epsilon)); + tt_i64_op(54, ==, add_laplace_noise(20, 0.75, delta_f, epsilon)); + tt_i64_op(100, ==, add_laplace_noise(20, 0.9, delta_f, epsilon)); + tt_i64_op(215, ==, add_laplace_noise(20, 0.99, delta_f, epsilon)); /* Test extreme values of signal with maximally negative values of noise * 1.0000000000000002 is the smallest number > 1 @@ -4143,54 +4143,54 @@ test_util_laplace(void *arg) */ const double noscale_df = 1.0, noscale_eps = 1.0; - tt_assert(INT64_MIN == + tt_i64_op(INT64_MIN, ==, add_laplace_noise(0, 0.0, noscale_df, noscale_eps)); /* is it clipped to INT64_MIN? */ - tt_assert(INT64_MIN == + tt_i64_op(INT64_MIN, ==, add_laplace_noise(-1, 0.0, noscale_df, noscale_eps)); - tt_assert(INT64_MIN == + tt_i64_op(INT64_MIN, ==, add_laplace_noise(INT64_MIN, 0.0, noscale_df, noscale_eps)); /* ... even when scaled? */ - tt_assert(INT64_MIN == + tt_i64_op(INT64_MIN, ==, add_laplace_noise(0, 0.0, delta_f, epsilon)); - tt_assert(INT64_MIN == + tt_i64_op(INT64_MIN, ==, add_laplace_noise(0, 0.0, DBL_MAX, 1)); - tt_assert(INT64_MIN == + tt_i64_op(INT64_MIN, ==, add_laplace_noise(INT64_MIN, 0.0, DBL_MAX, 1)); /* does it play nice with INT64_MAX? */ - tt_assert((INT64_MIN + INT64_MAX) == + tt_i64_op((INT64_MIN + INT64_MAX), ==, add_laplace_noise(INT64_MAX, 0.0, noscale_df, noscale_eps)); /* do near-zero fractional values work? */ const double min_dbl_error = 0.0000000000000002; - tt_assert(-35 == + tt_i64_op(-35, ==, add_laplace_noise(0, min_dbl_error, noscale_df, noscale_eps)); - tt_assert(INT64_MIN == + tt_i64_op(INT64_MIN, ==, add_laplace_noise(INT64_MIN, min_dbl_error, noscale_df, noscale_eps)); - tt_assert((-35 + INT64_MAX) == + tt_i64_op((-35 + INT64_MAX), ==, add_laplace_noise(INT64_MAX, min_dbl_error, noscale_df, noscale_eps)); - tt_assert(INT64_MIN == + tt_i64_op(INT64_MIN, ==, add_laplace_noise(0, min_dbl_error, DBL_MAX, 1)); - tt_assert((INT64_MAX + INT64_MIN) == + tt_i64_op((INT64_MAX + INT64_MIN), ==, add_laplace_noise(INT64_MAX, min_dbl_error, DBL_MAX, 1)); - tt_assert(INT64_MIN == + tt_i64_op(INT64_MIN, ==, add_laplace_noise(INT64_MIN, min_dbl_error, DBL_MAX, 1)); /* does it play nice with INT64_MAX? */ - tt_assert((INT64_MAX - 35) == + tt_i64_op((INT64_MAX - 35), ==, add_laplace_noise(INT64_MAX, min_dbl_error, noscale_df, noscale_eps)); @@ -4205,31 +4205,31 @@ test_util_laplace(void *arg) const double max_dbl_lt_one = 0.9999999999999998; /* do near-one fractional values work? */ - tt_assert(35 == + tt_i64_op(35, ==, add_laplace_noise(0, max_dbl_lt_one, noscale_df, noscale_eps)); /* is it clipped to INT64_MAX? */ - tt_assert(INT64_MAX == + tt_i64_op(INT64_MAX, ==, add_laplace_noise(INT64_MAX - 35, max_dbl_lt_one, noscale_df, noscale_eps)); - tt_assert(INT64_MAX == + tt_i64_op(INT64_MAX, ==, add_laplace_noise(INT64_MAX - 34, max_dbl_lt_one, noscale_df, noscale_eps)); - tt_assert(INT64_MAX == + tt_i64_op(INT64_MAX, ==, add_laplace_noise(INT64_MAX, max_dbl_lt_one, noscale_df, noscale_eps)); /* ... even when scaled? */ - tt_assert(INT64_MAX == + tt_i64_op(INT64_MAX, ==, add_laplace_noise(INT64_MAX, max_dbl_lt_one, delta_f, epsilon)); - tt_assert((INT64_MIN + INT64_MAX) == + tt_i64_op((INT64_MIN + INT64_MAX), ==, add_laplace_noise(INT64_MIN, max_dbl_lt_one, DBL_MAX, 1)); - tt_assert(INT64_MAX == + tt_i64_op(INT64_MAX, ==, add_laplace_noise(INT64_MAX, max_dbl_lt_one, DBL_MAX, 1)); /* does it play nice with INT64_MIN? */ - tt_assert((INT64_MIN + 35) == + tt_i64_op((INT64_MIN + 35), ==, add_laplace_noise(INT64_MIN, max_dbl_lt_one, noscale_df, noscale_eps));

1 0

← Newer
1
...
49
50
51
52
53
54
55
...
86
Older →

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

tor-commits November 2015