Re: [PATCH] gpg-interface: fix for gpgsm v2.3

From: Todd Zullinger <tmz@pobox.com>
Date: Feb 7, 2022, 16:38 UTC
Message-ID: <YgFK+F6Ks8FnN5Q6@pobox.com>
In-Reply-To: <20220207105240.dk443kcozynlonpp@fs>

Hi Fabien,

Fabian Stelzer wrote:

Show 5 quoted lines

> On 03.02.2022 15:01, Todd Zullinger wrote:
>> (To be clear, I mean "clever and wrong" in regard to my use
>> of the string list API, not anyone else's code.)
>
> string_list_split seems a bit like overkill.

I have little doubt that the string_list_split() method is far from ideal. :)

> I looked for sth like this but gpgs --help does not list it so i didn't dig
> deeper. I've checked the blame and it seems like this was introduced >19
> years ago. So i guess we can probably use this ^^

Indeed, the --with-colons output goes much further back in the GnuPG history than Git will ever have to care about.

Show 16 quoted lines

>>    --- a/t/lib-gpg.sh
>>    +++ b/t/lib-gpg.sh
>>    @@ -72,12 +72,10 @@ test_lazy_prereq GPGSM '
>>                    --passphrase-fd 0 --pinentry-mode loopback \
>>                    --import "$TEST_DIRECTORY"/lib-gpg/gpgsm_cert.p12 &&
>> 
>>    -	gpgsm --homedir "${GNUPGHOME}" -K |
>>    -	grep fingerprint: |
>>    -	cut -d" " -f4 |
>>    -	tr -d "\\n" >"${GNUPGHOME}/trustlist.txt" &&
>>    +	gpgsm --homedir "${GNUPGHOME}" -K --with-colons |
>>    +	awk -F ":" "/^fpr:/ {printf \"%s S relax\\n\", \$10}" \
>>    +		>"${GNUPGHOME}/trustlist.txt" &&
> 
> This does not quite work for me. It will add the fingerprint without the
> colons into the trustlist which is not valid :/

The colons are optional, and have been documented as such since cb1840720 ((Agent Configuration): New section., 2005-04-20). The text in the gpg-agent docs from GnuPG 2.2 say:

    Colons may optionally be used to separate the bytes of a
    fingerprint; this enables cutting and pasting the
    fingerprint from a key listing output.

Source: https://dev.gnupg.org/source/gnupg/browse/STABLE-BRANCH-2-2/doc/gpg-agent.texi;8021fe7670c79d5c698ec3fb600b02a9e5afb415$756?as=source&blame=off

How did it fail for you? It passes all the tests when I've run it against Fedora and RHEL-based hosts. If it's flaky on other systems, that would put a damper on doing it this way. Though it _should_ work.

[Note to myself] We don't just generate the key data, trustlist, etc. and store it in t/lib-gpg like we do with some other files per b41a36e635 (tests: create gpg homedir on the fly, 2014-12-12). That was because the gnupg home directory layout changed a bit between 2.0 and 2.1.

Thanks,

-- 
Todd