Commits on Source (70)
-
Daniel Kahn Gillmor authored
we work around by depending on gpg1 being available, which isn't a great solution. But i needed something to work for me right now.
b6a647b2 -
Daniel Kahn Gillmor authored033f9145
-
Daniel Kahn Gillmor authoredb7384e7d
-
Daniel Kahn Gillmor authored78d2fcca
-
Daniel Kahn Gillmor authoredcc5c2695
-
Daniel Kahn Gillmor authoreda0349d5d
-
Daniel Kahn Gillmor authored599e1baa
-
Daniel Kahn Gillmor authored875bf1c4
-
Daniel Kahn Gillmor authored4353b641
-
Daniel Kahn Gillmor authored7e177278
-
Daniel Kahn Gillmor authored536ff85a
-
Daniel Kahn Gillmor authored824eae93
-
Daniel Kahn Gillmor authored161e28d9
-
Daniel Kahn Gillmor authored13a3a242
-
Daniel Kahn Gillmor authoreda7d8fb41
-
Daniel Kahn Gillmor authorede2b02c52
-
Daniel Kahn Gillmor authored6d9cc776
-
Daniel Kahn Gillmor authoreda25f514b
-
Daniel Kahn Gillmor authoredc7b2462b
-
Daniel Kahn Gillmor authored
in gpg 2.1.x, subkey-to-ssh-agent doesn't work, because export-reset-subkey-passwd was dropped. We've now added a more performant agent-transfer tool to work in that scenario.
f126cecc -
Daniel Kahn Gillmor authored
We should be listing the public key when testing for capabilities; listing secret keys is slower and not necessary here.
4825d9ec -
Daniel Kahn Gillmor authoredd044a942
-
Daniel Kahn Gillmor authored47de7449
-
Daniel Kahn Gillmor authored
modern versions of gpg (>= 2.1.x) do not create secring.gpg, but we need secring.gpg for our use in add-servicename and revoke-servicename. If we can depend on gpg 2.1.13 and the --quick-revuid patch makes it in, it would be better to just use gpg's --quick-adduid and --quick-revuid, and maybe even to drop the add/revoke functionality from keytrans for simplicity.
339c8dd8 -
Daniel Kahn Gillmor authored48eece7f
-
Daniel Kahn Gillmor authored13ac8141
-
Daniel Kahn Gillmor authored37519ec6
-
Daniel Kahn Gillmor authored2a5a1e85
-
Daniel Kahn Gillmor authored
Use transferable key formats as inputs instead of assuming prior structure of a gpg homedir, which changes depending on the version of gpg you're using.
5cc3e56d -
Daniel Kahn Gillmor authored5fee2427
-
Daniel Kahn Gillmor authored17a195df
-
Daniel Kahn Gillmor authored
This avoids tickling the sockaddr_un.sun_path limits for me, since when i build in ~/src/monkeysphere/monkeysphere without a TMPDIR set, i see gpg-agent paths like: /home/dkg/src/monkeysphere/monkeysphere/tests/tmp/monkeyspheretest.VDc4Aek/testuser/.gnupg/S.gpg-agent which is quite close to the limit, and: /home/dkg/src/monkeysphere/monkeysphere/tests/tmp/monkeyspheretest.VDc4Aek/authentication/sphere/S.gpg-agent Which is actually over it if you include the terminating NUL Switching from tmp/monkeyspheretest.XXXXXXX to tmp/ms.XXX gives more breathing room (without fixing the underlying problem).
f227b33f -
Daniel Kahn Gillmor authored6682a8e2
-
Daniel Kahn Gillmor authored
some versions of gcc produce this warning, which is treated as an error due to our conservative defaults in Makefile: src/agent-transfer/main.c: In function ‘main’: src/agent-transfer/main.c:676:5: error: ignoring return value of ‘asprintf’, declared with attribute warn_unused_result [-Werror=unused-result] asprintf (&alt_comment, "GnuPG keygrip %s", args.keygrip); ^ cc1: all warnings being treated as errors this patch avoids the warning.
b756fd2e -
Daniel Kahn Gillmor authored
having src/share/common treated as an executable (commit ed10318d) was actually a terrible idea. In addition to causing "monkeysphere version" to print the version number twice, it meant that any invocation of a monkeysphere command that sourced src/share/common and had a first argument that happened to be a function name would accidentally invoke that function. This commit reverts that idea.
c75c7553 -
Daniel Kahn Gillmor authored0e339de4
-
Daniel Kahn Gillmor authoredef3e35cd
-
Daniel Kahn Gillmor authored
Standard installations should have the monkeysphere user's shell set to /bin/bash (or whatever is the preferred bash) already. But in some stranger situations that is not the case. This workaround should help fix those cases, and shouldn't cause any additional problems.
fe3bf727 -
Daniel Kahn Gillmor authored79e44fe1
-
2.1.15 appears to always emit the fingerprint lines in these cases, while 2.1.14 did not.
b1dd8fb1 -
d56774b1
-
Daniel Kahn Gillmor authored
clang noticed that there was a type mismatch if you believe that gpg_error_t is unsigned: ------- src/agent-transfer/main.c:677:13: error: comparison of unsigned expression < 0 is always false [-Werror,-Wtautological-compare] if (err < 0) { ~~~ ^ ~ 1 error generated. ------- Avoid this by using an independent int for the return of asprintf.
76c64fe8 -
Daniel Kahn Gillmor authored
on IRC, brethil reports that clang on OS X reports this failure when compiling agent-transfer: -------------- src/agent-transfer/main.c:467:27: error: incompatible pointer types passing 'struct sockaddr_un *' to parameter of type 'const struct sockaddr *' [-Werror,-Wincompatible-pointer-types] if (-1 == connect (ret, &sockaddr, sizeof(sockaddr))) { ^~~~~~~~~ /usr/include/sys/socket.h:583:41: note: passing argument to parameter here int connect(int, const struct sockaddr *, socklen_t) __DARWIN_ALIAS_C( connect); -------------- An explicit cast should address this concern. Ah, the joys of the BSD socket interface.
cb836a29 -
Daniel Kahn Gillmor authored
Apparently BSD sed and GNU sed treat -i differently, and there is no portable way to do in-place sed. This makes me sad and necessitates a fairly complicated workaround here. see also: https://unix.stackexchange.com/questions/92895/how-to-achieve-portability-with-sed-i-in-place-editing
d089d53b -
Daniel Kahn Gillmor authored
We want "make" to do all the file generation, so that no files are created while we're running "make install" as the superuser.
9fdef256 -
Daniel Kahn Gillmor authored
The installman target was generating files as well. Move those to their own build stage so that they get created cleanly.
ae0de274 -
Daniel Kahn Gillmor authored
when doing "make install" over an existing installation, the manpage links were failing because the link already existed.
7a39b6fc -
Daniel Kahn Gillmor authoredfb8e4756
-
Daniel Kahn Gillmor authored35c1fbf5
-
Daniel Kahn Gillmor authored40815ff5
-
Daniel Kahn Gillmor authored
When generating an OpenPGP certificate, include the v4 fingerprint in its hashed subpackets. This resolves https://bugs.debian.org/844971 when building against GnuPG 2.1.x
d086b1d3 -
Daniel Kahn Gillmor authored
https://bugs.debian.org/841208 This particular fix will only work for GnuPG 2.1.x or later, but should be safely ignored by older versions.
0828eba9 -
Daniel Kahn Gillmor authoredc29b216a
-
Daniel Kahn Gillmor authored23ee8532
-
Daniel Kahn Gillmor authored
Now that GnuPG reports fingerprints for all keys in GnuPG 2.1.16 (when using --with-colons), we need to tease out primary key fingerprints and to discard subkey fingerprints.
6d84352a -
Daniel Kahn Gillmor authored4af56ed4
-
Daniel Kahn Gillmor authored89657e03
-
Daniel Kahn Gillmor authoredd5ca4b9d
-
Daniel Kahn Gillmor authoredc0c005fa
-
a1e2bbc5
-
Daniel Kahn Gillmor authoreddfab82aa
-
-
Antoine Beaupré authored
-
Antoine Beaupré authored
-
Antoine Beaupré authored
-
-
50e53b5c
-
Daniel Kahn Gillmor authored225d7704
-
Daniel Kahn Gillmor authoredf427cb94
-
Daniel Kahn Gillmor authoredd21457f8
man/man1/agent-transfer.1
0 → 100644
src/agent-transfer/main.c
0 → 100644
This diff is collapsed.
src/agent-transfer/ssh-agent-proto.h
0 → 100644
src/share/common
100755 → 100644