- explicit conversions
- not all curves available for wolfSSL_CTX_set1_groups_list
- group funcs depend on HAVE_ECC
- `InitSuites` after `ssl->suites` has been set
- wolfSSL_CTX_set1_groups_list and wolfSSL_set1_groups_list should use wolfSSL_CTX_set1_groups and wolfSSL_set1_groups respectively because it converts to correct groups representation
- Change to using "SHA1" as main name for SHA1
If a client is:
- TLS 1.3 capable
- calls connect with wolfSSL_connect_TLSv13
- on an WOLFSSL object that allows downgrading
then the call to wolfSSL_connect should happen before changing state to HELLO_AGAIN. Otherwise wolfSSL_connect will assume that messages up to ServerHelloDone have been read (when in reality only ServerHello had been read).
Enable keying material for OpenVPN
in test.c: use use HEAP_HINT, not NULL in XMALLOC/XFREE calls;
fix a couple typos in aes_test() re WOLFSSL_ASYNC_CRYPT;
add various missing WOLFSSL_SMALL_STACK_STATIC qualifiers;
and streamline old smallstack refactor on-stack declarations declarations to use the much neater [1] construct.
Repology offers badges for wolfSSL's packaging status across all known
Linux distributions. [1] In Markdown documents, the HTML version with
a right-hand alignment uses up less space. It is added here.
The badge itself provides a hyperlink to more information [2]
including repositories in which wolfSSL is not yet represented.
[1] https://repology.org/project/wolfssl/badges
[2] https://repology.org/project/wolfssl/versions
during renegotiation).
This issue was brought to light by ZD 10911. When encryption is on (indicated
by the return value of IsEncryptionOn), DoHandShakeMsgType will finish up by
incrementing the input buffer index past the padding and MAC (if encrypt-then-
mac is enabled). In ProcessReply, if there are more messages to be read, the
index is decremented back before the padding and MAC. The issue arises when
ShrinkInputBuffer is called in between and copies data from the dynamic input
buffer to the static one. That function will get called with the index post-
increment, and thus the padding and MAC won't get copied into the static buffer,
which isn't what we want, since ProcessReply is going to decrement the index
since it thinks the padding and MAC are still there. This commit makes it so
the padding and MAC get included in the call to ShrinkInputBuffer when
encryption is on.
