Windows \r\n causes PEM object to fail parsing

[sergeyegerev]

Arduino/libraries/ESP8266WiFi/src/BearSSLHelpers.cpp

Line 359 in 7c2e071

if (!strcmp_P(pos[u].name, PSTR("CERTIFICATE")) || !strcmp_P(pos[u].name, PSTR("X509 CERTIFICATE"))) {

[sergeyegerev]

Hello.
I tried to use BearSSL::WiFiClientSecure for MQTT and used BearSSLX509List to parse X.509 certificate from AWS IoT. It comes in PEM format starting with -----BEGIN CERTIFICATE----- string.
At the above mentioned code line I've got pos[u].name value "CERTIFICATE-----" while the expected result is different. It looks like the name is parsed wrong.
Board version is 2.4.2.
Regards
Sergey

[sergeyegerev]

It seems like the issue is with line endings in the certificate: incorrect parsing occurs when lines end with CR/LF pairs (Windows line endings). Single LF (Unix) line ends are parsed OK.

[earlephilhower]

@sergeyegerev Thanks for the report. I'm able to reproduce using the validation example by changing to Windows EOL markers in a cert. This looks like it's a bug inside the semi-Forth text parser used to decode PEMs inside the BearSSL library itself. I've updated the bug title accordingly.