Change Details

In master, `gnupg_process_spawn` on Windows controls HANDLES accurately (while 2.4, 2.2 are *not*) ; It closes all HANDLES, when not specifically addressed. gpgtar supports `--status-fd` option. On Windows, it may be a HANDLE. gpgtar uses the HANDLE for its own status output, as well as passes it to subprocess of gpg which is invoked by `gnupg_process_spawn`. For the handling of the HANDLE of `--status-fd`, we have a bug which confuses the HANDLE and POSIX file descriptor. The bug doesn't expose a problem in 2.4 or 2.2, because its invocation of a process doesn't closes HANDLE but all HANDLES are inherit to subprocess. In master, `gnupg_process_spawn` does correctly, so, this bug is exposed.

In master, `gnupg_process_spawn` on Windows controls HANDLES accurately (while 2.4, 2.2 are *not*) ; It closes all HANDLES, when not specifically addressed. gpgtar supports `--status-fd` option. On Windows, it may be a HANDLE. gpgtar uses the HANDLE for its own status output, as well as passing it to subprocess of gpg which is invoked by `gnupg_process_spawn`. For the handling of the HANDLE of `--status-fd`, we have a bug which confuses the HANDLE and POSIX file descriptor. The bug doesn't expose a problem in 2.4 or 2.2, because its invocation of a process doesn't closes HANDLE but all HANDLES are inherit to subprocess. In master, `gnupg_process_spawn` does correctly, so, this bug is exposed.

In master, `gnupg_process_spawn` on Windows controls HANDLES accurately (while 2.4, 2.2 are *not*) ; It closes all HANDLES, when not specifically addressed. gpgtar supports `--status-fd` option. On Windows, it may be a HANDLE. gpgtar uses the HANDLE for its own status output, as well as passesing it to subprocess of gpg which is invoked by `gnupg_process_spawn`. For the handling of the HANDLE of `--status-fd`, we have a bug which confuses the HANDLE and POSIX file descriptor. The bug doesn't expose a problem in 2.4 or 2.2, because its invocation of a process doesn't closes HANDLE but all HANDLES are inherit to subprocess. In master, `gnupg_process_spawn` does correctly, so, this bug is exposed.