Nobody uses gpgtar for S/MIME

What happens if you call gpgtar with --utf8-strings --cms additionally to the other options? And what happens if you pipe the archive to gpgtar's stdin?

The original issue was to unclear to analyse and it was likely meanwhile fixed. For the other issue see the follow up ticket.

I mean this would also be solved if we did not use qiodevicedataprovider but pass the filenames directly to gpg for single files, too. (can't remember the ticket number) but I don't want to do that right now.

In T6526#177082, @ikloecker wrote:

The original issue was about creating an encrypted archive. This code doesn't use Qt anymore for writing the result file, but delegates this to gpgtar.

That sounds like a solid conclusion. I mean if errno is not set explicitly it is basically undefined which value it is, so maybe some other function set errno to no space left on device in that one case where it "worked".

The original issue was about creating an encrypted archive. This code doesn't use Qt anymore for writing the result file, but delegates this to gpgtar.

I've debugged Eva's problem and I think it's unrelated to the original problem, as it's specific to qt.

As what I see is similar as what Andre saw, I'll describe it here. Please check if this is relevant.
After the above mentioned Ticket was resolved, I tried the exact same encryption in Kleopatra on the same Test-VM.

That should be easy on Unix but on Windows we have the nul nul: and iirc also /dev/nul.

In T6556#175399, @werner wrote:

@iklocker: Which gpg bug to you mean?

@iklocker: Which gpg bug to you mean?

Thanks. For the record, done at https://lists.gnupg.org/pipermail/gnupg-users/2023-August/066692.html.

It may be better to open a separate issue for the issue in gpg, so that it's not overlooked/forgotten when the issue in gpgtar is fixed.

That is intentional. If we are able to remove a file we do it. Solution for you is easy: gpg .... -o - </dev/null >/dev/null

That is intentional. If we are able to remove a file we do it. Solution for you is easy: gpg .... -o - </dev/null >/dev/null

This looks like the same problem I encountered in Gentoo's Portage. To unlock the binary package signing key, Portage will run the equivalent of gpg --homedir ... --digest-algo ... --local-user ... --output /dev/null /dev/null. If unlocking fails (due to e.g. wrong password), /dev/null is removed: https://bugs.gentoo.org/912808

Yes, gpgtar emits a SUCCESS status. gpgme should probably check for this.

Works for me.

Normal priority to get the _1_ removed when no folder with the same name already exists in that location.

Strangely enough this does not happen on linux. Maybe related to the KMime changes we have there?

For S/MIME archives the output for e.g. testfolder.tar.gz.p7m is now named "testfolder.tar.gz_1_/testfolder" with the "_1_" even added if there is no "archive.tar.gz"

works. So gpgtar obviously knows about the filenames now, too

Closing since the problem doesn't seem to occur if the operation is canceled properly.

Sorry about that. I tested an old build which didn't call gpgme_cancel_async and therefore probably didn't properly close the channels. It seems to work if gpgme_cancel_async is called to cancel the operation.

This option is already used. Running pgrep -a gpg in a loop (and ignoring gpg-agent processes) I get:

Mo 26. Jun 11:29:11 CEST 2023
19111 gpgtar --batch --status-fd 60 --gpg-args --no-tty --gpg-args --charset=utf8 --gpg-args --enable-progress-filter --gpg-args --exit-on-status-write-error --gpg-args --display=:0 --gpg-args --ttyname=/dev/pts/37 --gpg-args --ttytype=xterm-256color --decrypt --directory /tmp/kleopatra-JqIiXu/src -- /home/ingo/dev/g10/src.tar.gpg
19112 gpg --batch --status-fd=60 --output - --decrypt --no-tty --charset=utf8 --enable-progress-filter --exit-on-status-write-error --display=:0 --ttyname=/dev/pts/37 --ttytype=xterm-256color -- /home/ingo/dev/g10/src.tar.gpg

Can you please test by adding --exit-on-status-write-error to the gpg invocation by gpgtar?

Should be fixed.

Due to the double fork in gpgme we won't get the exit code which gpgtar emits. Possible actions in a signal handler are also limited; in particular we can't use stdio or estream. The only option to print a status line would we by using write directly. However, this might mess with the libassuan buffering. Thus, it is not a good idea to pkill gpgtar. Same is true for gpg and gpgsm.

In the past this was done by --set-filename in libkleopatrarc-win32.desktop. But I am happy if we close this and focus on T6530.

Which only works if gpgtar actually knows the input file name (which it will once T6530: GPGME / QGpgME Extend Archivejobs to accept input / output from a filename is done and used).

Of course, those are different controllers. :-)

As I already had my testsetup open I recompiled with your change and I tested manually cancelling and letting it run into the disk full error. In both cases the temporary file was deleted and the job was cancelled :)

The processes should now be killed properly.

Please note that my test was not on an USB Device. I will keep this issue with your analysis and reopen a different one with my error and details on how to reproduce that one. Pretty sure it was disk full.