summary refs log tree commit diff
path: root/include/compat/netinet
diff options
context:
space:
mode:
authorJune McEnroe <june@causal.agency>2020-07-31 22:53:27 -0400
committerJune McEnroe <june@causal.agency>2020-07-31 23:10:51 -0400
commita6df11f2bbd2c9cdf4a8f16d93d8a56c8f41c68d (patch)
treeaa5fe01a11fa67f1dea2f3116f41a1684219542d /include/compat/netinet
parenttls_config: Replace constant with X509_get_default_cert_file() (diff)
downloadlibretls-a6df11f2bbd2c9cdf4a8f16d93d8a56c8f41c68d.tar.gz
libretls-a6df11f2bbd2c9cdf4a8f16d93d8a56c8f41c68d.zip
tls: Call SSL_CTX_set_default_verify_paths by default
This removes the hard dependency on a CA bundle file existing in the
default path (which seems to not be the case on Debian, for example),
but results in a subtle behaviour change: if the CA bundle file does not
exist, the CA directory will be used instead, rather than failing hard.

I believe the only reason libtls insists on loading a CA bundle file
itself is so that it can be sandboxed afterwards, given that a file is
loaded all at once while a directory is only loaded as needed. If the
default CA bundle file exists, SSL_CTX_set_default_verify_paths will
still immediately load it, so sandboxing will still work. If it doesn't
exist, then the CA directory will be used, which will work well for
unsandboxed applications, but will likely fail during verification as it
tries to search the directory. Either way, if the CA bundle file does
not exist, a sandboxed application will not work. Enabling the use of
the CA directory, however, will allow more unsandboxed applications to
work.

Finally, to restore the original behaviour, an application can call
tls_config_set_ca_file(3) with the path returned by
tls_default_ca_cert_file(3).
Diffstat (limited to 'include/compat/netinet')
0 files changed, 0 insertions, 0 deletions
lass='logsubject'>Rewrite glitch from new pngoJune McEnroe 2022-07-26Update Care with time-to-ID and piercingsJune McEnroe 2022-07-26Add -w to upJune McEnroe 2022-07-13Set push.autoSetupRemoteJune McEnroe 2022-07-08Remove TOURJune McEnroe There is not that much distinct stuff here anymore. 2022-07-03Add The Bone Shard EmperorJune McEnroe Suffers a little bit from middle book but I really enjoyed it. Read it faster than the first one too, despite its length. 2022-06-25Bump xterm font size to 12June McEnroe 2022-06-10Handle subshells (and functions) inside substitutionsJune McEnroe 2022-06-10Switch to jorts Install scriptJune McEnroe 2022-06-08Indicate if still reading or no resultsJune McEnroe 2022-06-08Add Maiden, Mother, CroneJune McEnroe Mixed bag like most collections of short stories. Some of them are pretty good. The author of the worst written story also has the worst written bio. 2022-06-05FIRST SHOW IN 2.5 YEARS BABEY!!!June McEnroe 2022-06-03Set line number on File linesJune McEnroe 2022-06-03Stop polling stdin after EOFJune McEnroe 2022-06-02Set TABSIZE=4June McEnroe Absolutely indiscriminately. 2022-06-02Do basic match highlightingJune McEnroe 2022-06-02Clean up parsing a littleJune McEnroe 2022-06-02Don't duplicate path stringJune McEnroe 2022-06-02Use stderr instead of /dev/tty, realloc buffer if lines too longJune McEnroe For some reason I haven't been able to figure out, trying to poll /dev/tty returns POLLNVAL (and this was using 100% CPU looping), but using stderr instead works fine. 2022-06-02Add initial working version of qfJune McEnroe 2022-05-29Set prompt for okshJune McEnroe