[FFmpeg-devel] [PATCH 0/5] [FATESERVER] Categorization and sorting improvements

[Timothy Gu]

On Sat, Jun 7, 2014 at 7:40 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
> On Sat, May 31, 2014 at 04:45:45PM -0700, Timothy Gu wrote:
>> After two days of hacking, I hereby present this patchset, which adds
>> support for querying, or "searching", a specific configuration; and
>> revamps the sorting system.
>>
>> With the new querying functioanlity, you can click on the CC, arch, or
>> OS to search for this same configuration, and you can click more of
>> those to narrow your search. Then, you can "clear all" search patterns.
>> Of course, this is not perfect: you cannot remove a single search
>> pattern if you have multiple, but this is a good starting point.
>>
>> I also improved the sorting algorithm. The current default, IMO, can
>> only be described as "bad". It is based on architecture, which usually
>> starts off with aarch64 or armv*. This is not useful at all for most
>> people, because most people want to see the latest -- and not ARM --
>> results. To use descending revision instead, I have to first add support
>> for descending default, which is impossible with the current sorting
>> system. Hence [PATCH 4/5]. Then, at [PATCH 5/5], I changed the default.
>>
>> Shortlog with comments:
>> [PATCH 1/5] fate.css: wrap long lines in log
>> Just a random style fix.
>>
>> [PATCH 2/5] index: move $params declaration to lsort()
>> Make $params local to lsort(). It doesn't really matter.
>>
>> [PATCH 3/5] Add support for querying a specific compiler, OS, and/or
>> [PATCH 4/5] index: use one parameter "sort" for both ascending and
>> [PATCH 5/5] index: improve sorting
>> See paragraphs above.
>
> btw, a bit off topic, but it might also be interresting to present
> the information in a 2d matrix where 2 of arch/os/compiler represent
> one axis and the 3rd the other axis

Yeah it would, something like Chromium's waterfall
(http://build.chromium.org/p/chromium/waterfall) is pretty visually
appealing. But I wonder if that would be feasible with the current
Perl CGI half-hackish approach of fateserver...

Timothy