2af81a7430481b3aec40f6a794db8cde4e1e18e4 introduced a bug where if the
`pgp-opportunistic-encrypt` attribute is set to `true` and a new message
is composed, aerc would crash.
This is because no recipients have been specified, so
`checkEncryption()` would fail early and *not* call `updateCrypto()`,
which would then cause a segfault in `SetEncrypt()` later on.

To avoid this, when the function would have returned early, it instead
*doesn't* set `c.encrypt` to `true` and still calls `updateCrypto()`.

Signed-off-by: witcher <witcher@wiredspace.de>
---
 widgets/compose.go | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/widgets/compose.go b/widgets/compose.go
index dc4719a..74493ac 100644
--- a/widgets/compose.go
+++ b/widgets/compose.go
@@ -1511,6 +1511,8 @@ func (c *Composer) checkEncryptionKeys(_ string) bool {
 			mk = append(mk, rcpt)
 		}
 	}
+
+	encrypt := true
 	switch {
 	case len(mk) > 0:
 		c.SetEncrypt(false)
@@ -1527,13 +1529,14 @@ func (c *Composer) checkEncryptionKeys(_ string) bool {
 			}
 		}
 		c.aerc.statusline.PushError(st)
-		return false
+		encrypt = false
 	case len(rcpts) == 0:
-		return false
+		encrypt = false
 	}
+
 	// If callbacks were registered, encrypt will be set when user removes
 	// recipients with missing keys
-	c.encrypt = true
+	c.encrypt = encrypt
 	err = c.updateCrypto()
 	if err != nil {
 		log.Warnf("failed update crypto: %v", err)
-- 
2.39.2

aerc/patches: SUCCESS in 4m38s

[pgp: fix segfault for opportunistic encryption][0] from [witcher][1]

[0]: https://lists.sr.ht/~rjarry/aerc-devel/patches/39455
[1]: witcher@wiredspace.de

✓ #950699 SUCCESS aerc/patches/openbsd.yml     https://builds.sr.ht/~rjarry/job/950699
✓ #950698 SUCCESS aerc/patches/alpine-edge.yml https://builds.sr.ht/~rjarry/job/950698

witcher, Mar 02, 2023 at 20:44:
> 2af81a7430481b3aec40f6a794db8cde4e1e18e4 introduced a bug where if the
> `pgp-opportunistic-encrypt` attribute is set to `true` and a new message
> is composed, aerc would crash.
> This is because no recipients have been specified, so
> `checkEncryption()` would fail early and *not* call `updateCrypto()`,
> which would then cause a segfault in `SetEncrypt()` later on.
>
> To avoid this, when the function would have returned early, it instead
> *doesn't* set `c.encrypt` to `true` and still calls `updateCrypto()`.
>
> Signed-off-by: witcher <witcher@wiredspace.de>

Acked-by: Robin-Jarry <robin@jarry.cc>

Applied. Thanks!