On 2 Oct 2021, at 20:44, Matthew Trescott via TortoiseSVN 
<[email protected]> wrote:

> I don't believe anything changed on its own since yesterday; TortoiseSVN was 
> still giving errors this morning. However, I completed the certificate chain 
> switch (certbot renew --preferred-chain 'ISRG Root X1' --force-renew) and 
> Tortoise is now satisfied, although my old Android now complains :)

That is great! It’s a fair trade-off. I’m pretty sure that SVN clients on old 
Android phones are extremely rare. I am definitely going to try the alternate 
chain. Sadly our ACME client (acme_tiny) does not support it yet, but we 
created a PR for it - it was rejected then, on the premise that it wouldn’t be 
necessary. But now that the DST expiration has finally happened, it turns out 
there are various corner cases where a sysop would want to choose the alternate 
chain.

> So indeed the problem is that Tortoise is only checking whether the highest 
> certificate in the chain is a trust anchor, and not considering that an 
> intermediary could be a trust anchor as well.

That is important information. thank you. I don’t know the TortoiseSVN source, 
but this MIGHT be caused by the way that the bundled OpenSSL is configured. The 
OpenSSL version is new enough, and this behavior is switchable. A patch was 
circulated on FreeBSD-security to change OpenSSL’s default behavior so that it 
should succeed at the first verified certificate and no longer bother looking 
at the expired DST root.

--- crypto/openssl/crypto/x509/x509_vpm.c.prev  2021-10-01
09:16:51.753533000 -0400
+++ crypto/openssl/crypto/x509/x509_vpm.c       2021-10-01
09:19:39.708106000 -0400
@@ -537,7 +537,7 @@
      "default",                 /* X509 default parameters */
      0,                         /* Check time */
      0,                         /* internal flags */
-     0,                         /* flags */
+     X509_V_FLAG_TRUSTED_FIRST, /* flags */
      0,                         /* purpose */
      0,                         /* trust */
      100,                       /* depth */

Perhaps (but this is absolutely above my pay grade) a similar configuration in 
TortoiseSVN’s build makes OpenSSL fail, and a change to OpenSSL’s configuration 
might be a structural solution for TortoiseSVN.

But first I am going to experiment with the short chain.

Thanks to all for your input!

Kind regards,
Walter Hop
Slik BV

-- 
You received this message because you are subscribed to the Google Groups 
"TortoiseSVN" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/tortoisesvn/61D2913C-83E8-4C00-9495-7BD4D2B4B318%40slik.eu.

Reply via email to