They weren't actually used since 7a35dad0

Bug 5809

Tom Fitzhenry authored 12 years ago and Nick Mathewson committed 12 years ago

bug4020

Signed-off-by: Tom Fitzhenry <tom@tom-fitzhenry.me.uk>

Our updateVersions script was being verbose, which might have made
sense back when I wrote it, but doesn't any more.

Final fix for bug 6599

Also remove some trailing whitespace.

Patch from maker; fixes bug 6024.

Mike Perry authored 12 years ago and Nick Mathewson committed 12 years ago

Tor2webMode is fingerprintable by hidden services through repeated
usage of the same three guard nodes for its rend and intro points.

Look at the address family of the preferred OR port rather than the
node.ipv6_preferred flag since the logic has changed with new
ClientUseIPv6 config option.

Fixes ticket 6884.

Asserts were hit by Tor2Web mode.

This affects both directory authorities and bridge authoritites.

Robert Ransom authored 12 years ago and Nick Mathewson committed 12 years ago

Now that crypto_pk_cmp_keys might return the result of tor_memcmp, there
is no guarantee that it will only return -1, 0, or 1.  (It currently does
only return -1, 0, or 1, but that's a lucky accident due to details of the
current implementation of tor_memcmp and the particular input given to it.)

Fortunately, none of crypto_pk_cmp_keys's callers rely on this behaviour,
so changing its documentation is sufficient.

Robert Ransom authored 12 years ago and Nick Mathewson committed 12 years ago

Fixes bug 4283; bugfix on r76
(Git commit 01aadefb).

Right-shifting negative values has implementation-defined behavior.
On all the platforms we work on right now, the behavior is to
sign-extend the input.  That isn't what we wanted in

    auth_type_val = (descriptor_cookie_tmp[16] >> 4) + 1;

Fix for 6861; bugfix on 0.2.1.5-alpha; reported pseudonymously.

The broken behavior didn't actually hurt anything, I think, since the
only way to get sign-extension to happen would be to have the top bit
of descriptor_cookie_tmp[16] set, which would make the value of
descriptor_cookie_tmp[16] >> 4 somewhere between 0b11111111 and
0b11111000 (that is, between -1 and -8).  So auth_type_val would be
between -7 and 0.  And the immediate next line does:

    if (auth_type_val < 1 || auth_type_val > 2) {

So the incorrectly computed auth_type_val would be rejected as
invalid, just as a correctly computed auth_type_val would be.

Still, this stuff shouldn't sit around the codebase.

We were doing (1<<p) to generate a flag at position p, but we should
have been doing (U64_LITERAL(1)<<p).

Fixes bug 6861; bugfix on 0.2.0.3-alpha; reported pseudonymously.

meejah authored 12 years ago and Nick Mathewson committed 12 years ago

add read_file_to_str_until_eof which is used by read_file_to_str
if the file happens to be a FIFO.

change file_status() to return FN_FILE if st_mode matches S_IFIFO
(on not-windows) so that init_key_from_file() will read from a FIFO.