Revert previous. It's too early to remove the guards just yet.

Remove HAVE_ED25519 now that we have ED25519 in LibreSSL

Handle spaces in key-p-tag correctly. Reported by Mischa Peters.

Make sure osmtpd_ltok_skip_hexdig doesn't derefences NULL

opensmtpd.h should be taken from standard include path

Remove assert. Shouldn't be in production code

Put pklen and tmp in #ifdef HAVE_ED25519

pkrawlen should only be used to test the length, not if the tag might be there or not

Apparently I broke RSA in previous commit

Add support for ed25519

Domains are case insensitive. Keep that in mind when comparing i- and d-tag

osmtpd_ltok_skip_sig_h_tag_value can be further out than end, because a header can contain a ';'

FWS after h=-tag value could lead to a end mismatch. Use osmtpd_ltok_skip_sig_h_tag_value for checking and original code for copying data

osmtpd_ltok_skip_key_h_tag_value should continue working on ptr, not prev

Use osmtpd_ltok_skip_sig_i_tag_value() instead of parsing it manually where not needed

Start the osmtpd_ltok_skip_sig_i_tag_value domain-name scan after '@'

osmtpd_ltok_skip_domain has obs-domain, which gives the wrong result. Use the appropriate osmtpd_ltok_skip_sig_d_tag_value

signature pkey also needs to be freed after we're done with it.

Move ar_data free to the end of the fuction, so that failed queries also get released

Free ar_data, as suggested by asr_run(3)

semicolon is part of qp-section/safe-char. Ignore it if we overflow

header->buf also needs to be freed

Add DKIM_UNKNOWN check for whitelines backlog as well

Rewrite osmtpd_ltok_skip_hyphenated_word. It didn't work quite right. While here remove unneeded #include <stdio.h>

Fix calculating subdomain matching for i tag