| Summary: | RequestTTY=no in config doesn't work if stdin is not a tty | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | Portable OpenSSH | Reporter: | Przemysław Pawełczyk <przemoc> | ||||||
| Component: | ssh | Assignee: | Damien Miller <djm> | ||||||
| Status: | CLOSED FIXED | ||||||||
| Severity: | normal | ||||||||
| Priority: | P2 | ||||||||
| Version: | -current | ||||||||
| Hardware: | All | ||||||||
| OS: | All | ||||||||
| Bug Depends on: | |||||||||
| Bug Blocks: | 1986 | ||||||||
| Attachments: |
|
||||||||
Created attachment 2171 [details]
move initialisation of tty_flag to after config parsing
The original patch seems no longer to apply. Here's a regenerated one.
patch applied. this will be in openssh-6.1 - thanks! Set all RESOLVED bugs to CLOSED with release of OpenSSH 7.1 |
Created attachment 2141 [details] Patch fixing RequestTTY=no that is not taken into account for non-tty stdin if set in config I've already reported the problem almost two weeks ago at #openssh, but apparently developers are available there only virtually (and I agree it's possibly not the best place for such activity). So this time I'm writing more proper bug report using bugzilla. I've noticed that RequestTTY option added in 8bf1740144a4 (May 2011) is not working 100% as intended (unless used in command-line). I checked the source code and it's because some tty stuff (touching options.request_tty) is performed before reading config file. I think that init log output, read per-user config, fill defaults and log reinit should be put just after buffer_init(...). Such fix has been tested and it seems to be working fine, but I may overlooked some (corner) cases, so please look at my solution in a cagey manner. Thanks.