Depending on timing we might get different errors (conn_reset, write failed)
and ignoring them all ends up being almost the same as just not checking
errors.
Rationale: if people want to disable RC4 but otherwise keep the default suite
list, it was cumbersome. Also, since it uses a global array,
ssl_list_ciphersuite() is not a convenient place. So the SSL modules look like
the best place, even if it means temporarily adding one SSL setting.
Reading the documentation of ssl_set_truncated_hmac() may give the impression
I changed the default for clients but I didn't, the old documentation was
wrong.
Goal is to test renegotiation better: we need more than one exchange for
server-initiated renego to work reliably (the previous hack for this wouldn't
work with non-blocking I/O and probably not with DTLS either).
Also check message termination in a semi-realistic way.
Will be useful for:
- detecting termination of messages by other means than connection close
- DTLS (can be seen as a special case of the above: datagram-oriented)
Found with Clang's `scan-build` tool.
ssl_mail_client.c does a dead store by assigning the return value of
base64_encode() to `len` and not using the value. This causes
scan-build to issue a warning.
Instead of storing the return value into `len`, store it to `ret`, since
base64_encode() returns a status code, not a length. Also check if the
return value is nonzero and print an error; this silences scan-build.
Found with Clang's `scan-build` tool.
The store to `ret` is not used, it's overwritten shortly after. Assign
the value of 1 at declaration time instead to silence scan-build.
Found with Clang's `scan-build` tool.
load_file() allocates memory to a char** parameter. It then tries to fread() a
file, and if that fails, frees the memory and returns to caller. However, the
char** is not reset to NULL, which causes a double-free error when the caller
later passes it to free().
