| 1 |
On Wed, Dec 14, 2016 at 7:56 AM, Mart Raudsepp <leio@g.o> wrote: |
| 2 |
> Ühel kenal päeval, K, 14.12.2016 kell 15:35, kirjutas Andrew Savchenko: |
| 3 |
>> On Wed, 14 Dec 2016 11:16:58 +0200 Mart Raudsepp wrote: |
| 4 |
>> > |
| 5 |
>> > Ühel kenal päeval, K, 14.12.2016 kell 13:08, kirjutas Sam Jorna: |
| 6 |
>> > > |
| 7 |
>> > > On Mon, Dec 12, 2016 at 02:35:28PM +1100, Sam Jorna wrote: |
| 8 |
>> > > > |
| 9 |
>> > > > |
| 10 |
>> > > > On Mon, Dec 12, 2016 at 09:34:21AM +0700, grozin@g.o |
| 11 |
>> > > > wrote: |
| 12 |
>> > > > > |
| 13 |
>> > > > > |
| 14 |
>> > > > > On Sun, 11 Dec 2016, Kristian Fiskerstrand wrote: |
| 15 |
>> > > > > > |
| 16 |
>> > > > > > |
| 17 |
>> > > > > > On 12/11/2016 03:13 PM, grozin@g.o wrote: |
| 18 |
>> > > > > > > |
| 19 |
>> > > > > > > |
| 20 |
>> > > > > > > gpg: signing failed: Inappropriate ioctl for device |
| 21 |
>> > > > > > this might indicate a want for export GPG_TTY=$(tty) |
| 22 |
>> > > > > I don't understand what has really happened. I removed my |
| 23 |
>> > > > > last |
| 24 |
>> > > > > commit, an |
| 25 |
>> > > > > attempt to commit it again failed with gpg: signing failed. |
| 26 |
>> > > > > Then |
| 27 |
>> > > > > I logged |
| 28 |
>> > > > > out of the box on which I have the git tree (I log in this |
| 29 |
>> > > > > box |
| 30 |
>> > > > > via ssh), |
| 31 |
>> > > > > and logged in again. After that the commit succeeded. |
| 32 |
>> > > > |
| 33 |
>> > > > I was also getting some odd issues with commit signing, though |
| 34 |
>> > > > it |
| 35 |
>> > > > seemed |
| 36 |
>> > > > to settle for me when I switched to pinentry-curses (since I |
| 37 |
>> > > > use |
| 38 |
>> > > > awesome), so I figured it was probably a local issue. Perhaps |
| 39 |
>> > > > there's a |
| 40 |
>> > > > wider problem here? |
| 41 |
>> > > |
| 42 |
>> > > If anyone else is getting this, it seems to be resolved by |
| 43 |
>> > > exporting |
| 44 |
>> > > GPG_TTY=$(tty) either immediately before attempting to sign or in |
| 45 |
>> > > your |
| 46 |
>> > > shell ~/.*rc file. |
| 47 |
>> > |
| 48 |
>> > I'd consider this a temporary workaround. The real issue would just |
| 49 |
>> > be |
| 50 |
>> > workarounded with this, which is nice to get something committed, |
| 51 |
>> > but |
| 52 |
>> > not so nice longterm. |
| 53 |
>> > I had similar issues, but it turned out some pinentry issues for me |
| 54 |
>> > iirc, so properly fixed by now and not needing such hackery |
| 55 |
>> > anymore. |
| 56 |
>> |
| 57 |
>> This is not a workaround, but officially recommended practice, from |
| 58 |
>> man gpg-agent: |
| 59 |
>> |
| 60 |
>> You should always add the following lines to your .bashrc or |
| 61 |
>> whatever initialization file is used for all shell invocations: |
| 62 |
>> |
| 63 |
>> GPG_TTY=$(tty) |
| 64 |
>> export GPG_TTY |
| 65 |
> |
| 66 |
> Then the packages or eselect pinentry or whatever should be taking care |
| 67 |
> of it, not have users have to mess with .bashrc to have stuff work. |
| 68 |
|
| 69 |
This is not practical. |
| 70 |
|
| 71 |
Adding it to the global /etc/bashrc is a bad idea. It would slow down |
| 72 |
every shell startup (fork/exec), even for users who do not actively |
| 73 |
use gpg (like root). |
| 74 |
|
| 75 |
Also, there is no way to know what shell each gpg user will be using. |
Archives