Bug in gpg 1.4.1/gpgme 1.0.2 - blocked while encrypting signed data with untrusted key

Marcus Brinkmann marcus.brinkmann at ruhr-uni-bochum.de
Mon Aug 29 14:10:50 CEST 2005

At Wed, 17 Aug 2005 20:37:42 +0200,
Stéphane Corthésy wrote:
> [GNUPG:] GET_BOOL untrusted_key.override

GPGME was not prepared to deal with such unexpected questions on the
command fd.  This was hopefully fixed a couple of days ago in the HEAD:

2005-08-26  Marcus Brinkmann  <marcus at g10code.de>

        * rungpg.c (command_handler): Use _gpgme_io_write instead of write.

        * edit.c (command_handler): Do not depend on PROCESSED being

        * engine.h (engine_command_handler_t): Add new argument processed.
        * ops.h (_gpgme_passphrase_command_handler_internal): Rename
        prototype to ...
        (_gpgme_passphrase_command_handler): ... this one.
        * passphrase.c (_gpgme_passphrase_command_handler_internal):
        Rename to ...
        (_gpgme_passphrase_command_handler): ... this one.
        * edit.c (command_handler): Add new argument processed.  Remove
        local variable with the same name.  Always return processed as
        * rungpg.c (command_handler): Send a newline character if the
        handler did not.

If you can confirm that it works, I can push the fix to the 1.0 branch.

> I hope it's been fixed in gpg 1.4.2/gpgme 1.1 (when will they be  
> released?), as it prevents the use of gpgme with the 'trust all keys'  
> option.

GPGME 1.1 will be released "soonish".


More information about the Gnupg-devel mailing list