Bug 1479: GnuPG curl-shim TCP half-close harms HTTP interop

David Shaw dshaw at jabberwocky.com
Sat Mar 2 02:19:03 CET 2013

On Mar 1, 2013, at 6:56 PM, Phil Pennock <gnupg-devel at spodhuis.org> wrote:

> On 2013-03-01 at 17:46 -0500, David Shaw wrote:
>> I think you are confused.  The structure ctx is a struct curl_writer_ctx, not a struct CURL.  It's defined in ksutil.h.
> Crap, skimmed too quickly, sorry.
>> Why did you put "KEY" in front of the key ID of the key?  GPG doesn't do that.
> Because I didn't spot documentation and didn't read the source in depth,
> I skimmed to get the minimal possible to try to figure out why, in
> normal use without invoking the helpers manually, the verbose/debug
> output was both showing a key and reporting that no key is found, for
> both myself and the other person tracking down the gnupg/keyserver
> interop issues.
>> Aside from those two points, this works for me.  You snipped most of the output, so the best guess I can give you is that for some reason you're missing the "-----END PGP PUBLIC KEY BLOCK-----".
> The keys are retrieved, it's just that "--keyserver-options
> verbose,debug" erroneously reports that there's no key.

While I believe you are seeing this, I'm not, and you're reporting a bug (keyserver fetches not working at all) that I daresay would have been noticed a long time ago.

So, is your setup unusual in any way?  Are you going through a proxy?  What platform are you running on?  Which (exact) version of the GPG code are you running?  Does it happen when building with curl (yes, I understand you saw it when verifying a bug that only applies to curl-shim).  Does it work when you don't apply the shutdown patch?


More information about the Gnupg-devel mailing list