Re: Lintian and Dpkg's :any multiarch acceptor

2021-11-04 Thread Debian/GNU

On 11/3/21 23:34, Felix Lechner wrote:

1. Did anyone find the latest Lintian versions (2.109.0 and up)
confusing as to whether the :any should be included? The material you
would have encountered includes both the context offered by Lintian
(the extra information after the tag) and any relevant tag
descriptions.


being the one who initially triggered #995981 by blindly following 
advice of lintian (yes i know: *never* blindly follow lintian's advise), 
i think i can say that i indeed found the description of the 
"rules-require-build-prerequisite" tag highly confusing.


in general i think tag descriptions should ony use "machine-parsable 
advice" if the machine-parsable is meant to be used as-is, and use human 
language in all other cases.
that is: rather than using a string like "python3:any | python3-all:any 
| python3-dev:any | python3-all-dev:any | dh-sequence-python3" (which 
looks very much like code) use someting like "the package should 
probably build depend on one of python3-dev:any or dh-sequence-python3 
or [...]".


the point i'm trying to make is not about the correctness of the advice 
itself, but that an explanation that looks like a code-example is 
probably going to be understood as such.
so if the text is not to be copy and pasted into code, it shouldn't look 
like tit could.


fmdst
IOhannes


OpenPGP_signature
Description: OpenPGP digital signature


Lintian and Dpkg's :any multiarch acceptor

2021-11-03 Thread Felix Lechner
Hi,

For a brief time between October 1 and October 15, Lintian gave
potentially confusing advice on some build prerequisites. [1]

The :any multiarch acceptor—a rarely used feature some other tools
call the "muliarch qualifier"—was originally not implemented at all
[2] and then implemented incorrectly. [3] Many people do not even know
about the feature. To my knowledge it works now.

Here are two questions:

1. Did anyone find the latest Lintian versions (2.109.0 and up)
confusing as to whether the :any should be included? The material you
would have encountered includes both the context offered by Lintian
(the extra information after the tag) and any relevant tag
descriptions.

2. Should Lintian issue any advice when it sees the :any multiarch
acceptor? If so, for which packages? It might allow maintainers to
undo erroneous advice they may have been given, although many folks
use the feature legitimately, as well.

Thank you!

Kind regards
Felix Lechner

[1] The affected versions were 2.107.0 and 2.108.0.
[2] https://bugs.debian.org/994902
[3] https://bugs.debian.org/995981