Fixed in GnuPG 2.3.3.

I should have explained the context.
No, there is no discussion about this in the WG.

Is that really required? Should we wait what the conlusion of the WG will be?

I'm reading RFC5297, which says:

SIV can be used as a drop-in replacement for any specification that uses [RFC3394] or [RFC3217], including the aforementioned use. It is a more general purpose solution as it allows for associated data to be specified.

Fix for this issue landed RNP master, and will be included to the RNP v0.16.0 release.
Within fix:

• new keys will be generated with correctly tweaked bits
• using secret key with non-tweaked bits would issue a warning
• CLI command --edit-key [--check-cv25519-bits | --fix-cv25519-bits] added, allowing to fix older key

As long as we can't replicate this, it does not make sense to keep this bug open. Please re-open it if you run into it again in a replicatable way.

Use of version 5 format for Ed448/X448 was pushed by rG86cb04a23d2b: gpg: Ed448 and X448 are only for v5 (for subkey)..

Bug in creating such a blob is fixed in rG08a3a4db27dc: kbx: A 20 byte fingerprint is right filled in version 2 blob..

Thanks for clarification, indeed attempt to decrypt data returns an error afterwards.

Well, while importing you get the warning:

Yes, for migration from GnuPG 2.0 reasons, a batch import delays the key checking (i.e. converting from OpenPGP to GnuPG internal format) to the first use. Thus you don't see an error immediately. But if you encrypt something , you won't be able to decrypt it again:

Thanks, Werner.
During further work on this got another issue:

Thanks. I meanwhile pushed a fix to 2.3 so that a warning is shown if the low bits are set.

Thanks for the replies, this makes things clear. We'll update RNP to correctly set/unset those bits while saving a generated secret key and a way to fix up previously generated keys.

Right, as long as there is only one format in widespread use (based on a long existing 4880bis draft) only this format should go over the wire.
Thus, it is a matter how the key is exported. In cryptography you should never have several options - one clearly defined format is what you want. We have had enough trouble with PGP5 peculiarities but in that case their implementation had more users and thus GnuPG had to work around it. Not good, but there was no standard at all at this time.

@onickolay No sorry needed. It was me, who cannot answer promptly.

@gniibe sorry for pinging, but this issue gets attention as TB users (with RNP OpenPGP backend) cannot import to GnuPG EdDSA secret key which was generated by RNP since it doesn't tweak bits when storing or exporting a secret key.
Should we update RNP to tweak those bits during storage to be more compatible (given that those bits doesn't make any difference)?

Do I correctly understand that issue will be resolved on GnuPG side by tweaking key bits before private-key import/and/or/operations?

We should not support a different OID or representation of 22519 which will only lead to incompatibilities and trouble existing users. 25519 is in too widespread use than to allow for any changes.

@werner isn't it used just for the public key? The secret x25519 key, exported by GnuPG, looks as following (in the way it is stored in file):

We invented the 0x40 compression flag to declare that as native curve point format. With the introduction of 448 things got more complicated due to the new IETF statdards for this curev. This is the reason for @gniibe's proposal for a Simple Octet String (SOS) as a new data type in OpenPGP.

Investigated it more, and it looks problem is not in incorrect endianness. Exporting x25519 secret subkey from the GnuPG showed up that we still need to change byte order.
After some experiments I ended up with the following self-explaining code piece, which makes RNP-generated keys to work with GnuPG for import:

repeat:
    if (botan_privkey_create(&pr_key, "Curve25519", "", rng_handle(rng))) {
        goto end;
    }
    /* botan returns key in little-endian, while mpi is big-endian */
    if (botan_privkey_x25519_get_privkey(pr_key, keyle.data())) {
        goto end;
    }
    if ((keyle[31] != 0x45) || (keyle[0] != 0x40)) {
        botan_privkey_destroy(pr_key);
        goto repeat;
    }
    if (botan_privkey_export_pubkey(&pu_key, pr_key)) {
        goto end;
    }

Thanks for investigations! Indeed, we do change byte order when storing/loading private key, as MPI should be big-endian, while curve25519 private key is little endian.
Do I correctly understand that we should store it in the MPI as it is (like with Ed25519)? It would be nice to clarify that in the RFC draft.
Another thing is that in my test even if byte order is not reversed in the secret key (including the attached test key), GnuPG still asks for password, reporting "error sending to agent: Bad passphrase".

The problem here appears to be that the "MPI" of the curve25519 secret key is not actually a standard-issue big-endian OpenPGP MPI -- it's an opaque bytestring expected to be passed to the underlying "native" implementation of x25519, in the same way that the secret key is handled for Ed25519.

investigating the subkey in python:

looks to me like you've got the byte ordering of the Curve25519 secret subkey reversed from the way that GnuPG expects it.

fwiw, gpg-agent complains that the keys don't match:

Due to tax issues, we can't accept a donation as return on service. However, we will fix bugs anyway if possible,

The curve is not defined to be used for ECDH (encryption); in fact it should in general only be used with the EdDSA
algorithm. You need to use "Key-Type: eddsa". Note that the EdDSA signing algorithm is different than the commonly used ECDSA signing algorithm.

Thanks for the quick response Werner. I knew I could use it with quick-gen-key and I’ve updated my config file to have it as default.
But, just for my understanding, is there a reason ed25519 cannot be used with full-gen-key and gen-key in batch mode?

You can't use ecdh with ed25519.

I just realized that my example is incorrect. It doesn't make sense to support multiple issuer subpackets on self signatures. But it is useful to do so on binary signatures and third-party certifications. Here's a better example, which gpg correctly supports. As such, this issue should be closed. Sorry for the noise.

I implemented subkey collapsing in 2.3. It is enabled by default but you can disable it it with

OpenPGP (RFC-4880) requires support for 3DES and SHA-1 thus you can't disable them. However, they are not used in practice because the key preference guarantee the use of more modern algorithms,

Since it was handled in T4908, this task is merged into that.

@mbrinkers : I think that it was fixed in GnuPG 2.2.21 by T4908: ECDH with AES-128 decryption failure when fully padded.
It was unfortunate that this bug report didn't work to solve problem, with malformed data and discussion went to unrelated thing.

I have run into an interoperability issue between BouncyCastle PGP (Java) library and gpg which seems to caused by key obfuscation.

In the SOS branch, rG1c4291c3951d: ecc-sos: Add special leading zero octet removal. should be reverted.
Instead, the S_KEY should be fixed up in read_key_file in findkey.c,
and merge_lists in protect.c.
(Then, no need to be fixed up in extract_private_key.)

Exactly same problem is there in libgcrypt.
In the definitions of curves, it uses negative constant internally in some specific places, but for other parts, we have same problems.

I should concentrate the case of ECC, in particular, ECC with modern curves.
Removing leading zero from RSA/ECC/ELGamal assuming unsigned integer would result more work.

In libgcrypt, we have another problem of GCRYSEXP_FMT_ADVANCED formatting, which is used by gpg-agent of GnuPG 2.3 with name-value list.

Confusingly, in the SSH specification, it is signed MPI.
See RFC4251, for the definition of "mpint": https://tools.ietf.org/html/rfc4251#page-8

There are more places for clean up in GnuPG.
While "MPI" in OpenPGP specification is based on unsigned integer, the default "MPI" handling of GnuPG/Libgcrypt is signed. This difference matters internally.
Formatting by "%m" with libgcrypt, it may result prefixed by 0x00 (so that it represents unsigned value, even if scanned as signed).
And because of this, existing private keys in private-keys-v1.d may have this leading zero-byte.
But the counting bits don't count this byte.

Important interoperability issue:
OpenPGP implementations should implement:

• Recovery of leading zero octets for Ed25519 key handling (secret part) and Ed25519 signature

Better to paste directly:

# SOS representation
#
# Initially, it was intended as "Simply, Octet String", but 
# it is actually "Strange" Octet String.
#

I wrote this:

Sorry, I don't know what kind of sample data that is. The reference keys have been provided by the RFC6637 author and are part of GnuPG's test suite; see (gnupg/tests/openpgp/samplekeys/ecc-sample-*).

I can't find any places where it is interpreted as signed integer.

It seems that the reference to PKCS#5 is correct. It is an issue of how to describe the case of more than 8-byte padding in OpenPGP.

Your example data is malformed, I suppose.

I also don't think that key size obfuscation is useful, after all the preferences of the key demand a certain key size.

It is easy to explain:

That means that the GnuPG Backend does not work. I do not think that the office update is the reason, me and others use GpgOL with the most recent versions of Office Pro Plus without issue.
Have you possibly modified you gnupg config files? If there is a bad value in there it would result in such an error.

Very few OpenPGP data signatures have an expiration time either, fwiw. I have never actually seen one in the wild, and no one that i know uses --ask-sig-expire or --default-sig-expire (it shows up in the cupt test suite and the apt test suite, but doesn't appear to be actually used by anything).

CMS signatures do not have a expiration time. Further the meaning of the expiration time of one of the certificates also depends on the validation model (shell or chain); thus a one-to-one relationship between these times is not possible.

For pinpadtest.py, you need to offer an option --add (adding dummy byte), when you are using Cherry ST-2xxx.

For pinpadtest.py, you need to offer an option --add (adding dummy byte), when you are using Cherry ST-2xxx.

It is not supported, by CCID protocol itself. So, it is not supported by scdaemon, and by any of card readers (which I know of), either.

It is not supported, by CCID protocol itself. So, it is not supported by scdaemon, and by any of card readers (which I know of), either.

Note that rGd3f5d8544fdb needs to be backported to 2.2 but we will wait until we have better tested it.

I am in charge of editing the current OpenPGP draft, so I will for sure keep an eye on that issue. If would appreciate if you can post your report also to openpgp at ietf org.

I am trying to reproduce your problem with my 3.3 card using my TTXS card reader.

I use the internal driver.

Are you using pcscd (is that process running) or the internal driver.? Please try the latter if you are not already using it.

It was rG07250279e7ec: * keyedit.c (keyedit_menu): Invisible alias "passwd" as "password". in 2004, which set default to rfc2440-text behavior.
And in 2007, the commit rGb550330067b6: * gpg.c (main): Disable --rfc2440-text and --force-v3-sigs by default. changed the default to no-rfc2440-text.

I see. Thanks for your explanation.