…value)

…el mutations)

TODOs:
 - correctly evaluate any sub-term in MakeMessage
 - modify evaluate for terms and replace found payloads_0 by payloads

…igate zoo

…ests

…ture, fully tested (in seeds.rs)

…e sup-terms of non-symbolic, fix many bugs, test to investigate mutation failures

TODO: clear understanding of which mutations and replacement can fail

… Countable trait + various fixes

…ion symbolc like fn_append

…ba to get picked)

…oir sampling

…orked as it does not well handle multiple payloads reaplcements inducing indices shifts

…ever MakeMEssage a term that is not fullly symbolic (i.e., that has some payload in sub-terms), many tests!

…tations.rs)

…ayloads

…hPayloads

This reverts commit 457fb47.

…cements. Test test_replace_bitstring_multiple is more complete and now passes!

… to have a fine-grained control about how it is encoded/read.

Vec<T> is encoded as the series of the encoding of the items (without length prefix). Wrapper(T) is encoded by adding a length prefix whose teh exact format depends on T/Wrapper. For bit-level mutations, this also improves things: (i) mutate Wrapper can mutate the length of the list, mutate Vec<T> can alter the structure of the encoded vector and (ii) we can more easily find where to replace payloads when scanning sub-terms of the form Vec<T>.

- impacted/new types: Compressions, CipherSuites, ClientExtensions, ServerExtensions, HelloRetryExtensions, CertificateEntries.
- rustls::msgs:message.rs now uses those new encode/read implems for those types
- adding to the signature symbols/functions like fn_cipher_suites_make that tranform Vec<T> into Wrapper(T)
- modify seeds and vulns accordingly
- tests pass, including the new `test_replace_bitstring_multiple`

# Conflicts:
#	puffin/Cargo.toml
#	tlspuffin/Cargo.toml
#	tlspuffin/src/tcp/mod.rs

…tiple

During PUT OpenSSL initialization a confusing message about the
non-availability of clear mode is logged, but only when clear mode is NOT
selected. In fact, the OpenSSL PUT only implements clear mode rather than a
full reset of the agents.

This is confusing on two levels:
- the message hints that a full reset of OpenSSL agents is always performed but
  they are, on the contrary, always soft reset.
- the process continues with the soft reset without making it clear to the user
  that the option they have selected won't be applied.

We implement the non-clear mode in the OpenSSL PUT to get rid of the message
altogether.

Fixes: #154

…gful data

The `tlspuffin execute` command exits with 0 or 1 to signal if there are
potentially more inputs that were not executed. This has been used for batch
executions during coverage analysis.

Numerous tools, including the running shell, associate a single meaning to
non-zero exit code: the command failed to execute in the way intended by the
caller. Depending on how they are configured, these tools might behave
differently when they detect a non-zero return code (e.g. bash "set -e").

To avoid any confusion, we instead log the total number of inputs, so that
calling scripts can still retrieve it without relying on the exit code.