[FFmpeg-devel] core infrastructure badge for FFmpeg

[Ganesh Ajjanagadde]

06.07.2016, 08:51, "Jean-Baptiste Kempf" <jb at videolan.org>:
>  On 06 Jul, Ganesh Ajjanagadde wrote :
>>   > But, you could at least be a bit more truthful when filling it:
>>   >  - the buildsystem is not common tools, since you have your own
>>   >    configure (it's a SUGGESTED thing anyway)
>>
>>   As far as I know, shell script is a common tool available on POSIX platforms.
>
>  This is not what they mean:
>  They mention "For example, Maven, Ant, cmake, the autotools, make, or
>  rake". Else, they would mention shell script, since it's the base of
>  many of those tools.
>  A custom script is not a common build system tool.

How configure.ac that gets fed into autotools etc is acceptable and configure is not is beyond me.
As for their meaning, I don't know.

>>   >  - the new functionnality testing is only done in libavcodec,
>>
>>   I did not assert to the contrary, but it is certainly an "informal" policy per the dev docs.
>>   I did acknowledge the lack of tests in certain places such as libavfilter.
>
>  Not only libavfilter.

So you want me to change from "libavfilter" to "everything except libavcodec"?
Note that it is anyway marked as "not met".

>>   >  - half of the links given are over github, which is not FLOSS and is
>>   >    just a mirror... And other on ffmpeg.org
>>
>>   So what? There is no requirement that the links must be posted on FLOSS sites,
>>   whatever that means which you conveniently leave unspecified.
>>
>>   The purpose of the mirror here is just as a reference point, and can always be changed.
>>   There is some lack of consistency in link usage, I am going to change it.
>
>  I just mean that this is not coherent. Especially on something that is
>  not controlled by you.

Changed, all gratuitous github links have been amended. Thanks.

>>   >  - you do not use SEMVER (it's a SUGGESTED improvement too)
>>
>>   I thought we did, since that is what I assumed the chicanery regarding major, minor, etc was all about.
>>   Will amend.
>
>  SEMVER is stupid anyway.

Amended.

>>   >  - everything related to external users should be N/A and not "met",
>>   >    (I doubt they mean trac here)
>>
>>   Please be more explicit here. What is wrong with trac as a link for a bug tracker?
>
>  Good cryptographic practices, questions 8 and 9 should be "N/A", not
>  "met": you don't store users credential, as you say in the comment.

So you are saying that regarding RNG, Q9 is referring to internal use by the project infrastructure?
Otherwise I don't see how it is "N/A"; it is either met or not.

>>   >  - where is written the policy that coverity must be run on each release?
>>
>>   No policy, but Michael did say that "we do run coverity around the release time generally".
>>   From my experience over the last 6 months, Coverity was run at least once before each release.
>>   If there was an instance where it was not, sure, I will remove it.
>
>  "said" and "policy" are different things, notably for MUST items.

Ok, changed to "not met" for now.

>  --
>  Jean-Baptiste Kempf
>  http://www.jbkempf.com/ - +33 672 704 734
>  Sent from my Electronic Device
>  _______________________________________________
>  ffmpeg-devel mailing list
>  ffmpeg-devel at ffmpeg.org
>  http://ffmpeg.org/mailman/listinfo/ffmpeg-devel