Change Details

In libgpg-error (also in libassuan, and gnupg 2.4.*), spawning a process with "NUL", it actually has not been working well. `w32_open_null` function opens "NUL", but it is not inheritable to child process (currently, until we will fix). Fixing this bug will introduce a regression (or two) in GnuPG, specifically, in `gpg-connect-agent`; `gpg-connect-agent` uses `gnupg_isatty` to distinguish if it's user's console session or not. On Windows, `isatty` returns TRUE for "NUL" (or any character devices), which is not expected. Thus, some sort of bug compatibility support is needed. My proposal is use of new flag `GPGRT_PROCESS_IO_NUL` means enabling "NUL". Applications which require the behavior of having "NUL" for stdin/stdout/stderr should specify the flag `GPGRT_PROCESS_IO_NUL`. Otherwise, it's backward compatible (invalid handle in a child process).

In libgpg-error (also in libassuan, and gnupg 2.4.*), spawning a process with "NUL", it actually has not been working well. `w32_open_null` function opens "NUL", but it is not inheritable to child process (currently, until we will fix). Fixing this bug will introduce a regression (or two) in GnuPG, specifically, in `gpg-connect-agent`; `gpg-connect-agent` uses `gnupg_isatty` to distinguish if it's user's console session or not. On Windows, `isatty` returns TRUE for "NUL" (or any character devices), which is not expected. Thus, some sort of bug compatibility support is needed. My proposal is use of new flag `GPGRT_PROCESS_STDIO_NUL` means enabling "NUL". Applications which require the behavior of having "NUL" for stdin/stdout/stderr should specify the flag `GPGRT_PROCESS_STDIO_NUL`. Otherwise, it's backward compatible (invalid handle in a child process).

In libgpg-error (also in libassuan, and gnupg 2.4.*), spawning a process with "NUL", it actually has not been working well. `w32_open_null` function opens "NUL", but it is not inheritable to child process (currently, until we will fix). Fixing this bug will introduce a regression (or two) in GnuPG, specifically, in `gpg-connect-agent`; `gpg-connect-agent` uses `gnupg_isatty` to distinguish if it's user's console session or not. On Windows, `isatty` returns TRUE for "NUL" (or any character devices), which is not expected. Thus, some sort of bug compatibility support is needed. My proposal is use of new flag `GPGRT_PROCESS_STDIO_NUL` means enabling "NUL". Applications which require the behavior of having "NUL" for stdin/stdout/stderr should specify the flag `GPGRT_PROCESS_STDIO_NUL`. Otherwise, it's backward compatible (invalid handle in a child process).