W3C home > Mailing lists > Public > public-aria-editors@w3.org > November 2016

Re: APG build fail

From: Michael Cooper <cooper@w3.org>
Date: Mon, 14 Nov 2016 12:56:45 -0500
To: "Michiel Bijl (list)" <michiel.list@moiety.me>
Cc: Matt King <a11ythinker@gmail.com>, ARIA Editors <public-aria-editors@w3.org>
Message-ID: <d735f5ab-cd2a-b801-9c86-fba40fab5783@w3.org>
I really don't know how Bikeshed works, so can't answer the question. We 
have a degree of investment in Respec now, with various extension 
functions we use to tie the ARIA stuff together. So it's not easy to say 
we should just switch. But if you think Bikeshed is a viable alternative 
and worth the investment in transitioning, I welcome a proposal. Michael

On 09/11/2016 5:15 AM, Michiel Bijl (list) wrote:
> Wouldn’t something like Bikeshed allow for that?
> —Michiel
>> On 8 Nov 2016, at 22:17, Michael Cooper <cooper@w3.org 
>> <mailto:cooper@w3.org>> wrote:
>> It's gotta access the files from somewhere. AFAIK the spec converter 
>> service only has a web interface. I'd love to figure out how to 
>> install the spec converter, or a comparable tool, directly within the 
>> travis build so it could use copies of the files that are available 
>> within the build process without requiring access to them over 
>> rawgit. But at the moment I don't know how to do that. Michael
>> On 08/11/2016 2:05 PM, Michiel Bijl (list) wrote:
>>> The erros seem to be with Rawgit, why does the travis build even 
>>> need Rawgit?
>>> —Michiel
>>>> On 8 Nov 2016, at 17:09, Matt King <a11ythinker@gmail.com 
>>>> <mailto:a11ythinker@gmail.com>> wrote:
>>>> OH, not with travis … totally manual, downloading index.html from 
>>>> rawgit.
>>>> For now, I have 2 clones in sibling subdirectories on my system, 
>>>> one with just the gh-pages branch and a main clone with other 
>>>> branches. So, copying examples from master to gh-pages is trivial.
>>>> This is not to say that I don’t really miss the automatic build …. 
>>>> It is just survival technique.
>>>> On the the topic of respect,  I doubt it has anything to do with 
>>>> the automatic build fails, but the need for browser refresh may be 
>>>> related to the same issue that is effecting the AX tree. I think 
>>>> respect should do all processing in a node that is not displayed. 
>>>> While the processing is happening, it should just display a single 
>>>> node saying “Document being formatted by respect.”. Then, when 
>>>> processing is done, change the display property of the complete 
>>>> document. This would be much more screen reader friendly.
>>>> Matt
>>>> *From:*Michiel Bijl [mailto:michiel.list@moiety.me]
>>>> *Sent:*Tuesday, November 8, 2016 8:55 AM
>>>> *To:*Matt King <a11ythinker@gmail.com <mailto:a11ythinker@gmail.com>>
>>>> *Cc:*Michael Cooper <cooper@w3.org <mailto:cooper@w3.org>>; ARIA 
>>>> Editors <public-aria-editors@w3.org 
>>>> <mailto:public-aria-editors@w3.org>>
>>>> *Subject:*Re: APG build fail
>>>> How do you the manual build? Because I don't see the build button 
>>>> Michael talked about.
>>>> —Michiel
>>>> On 8 Nov 2016, at 16:47, Matt King <a11ythinker@gmail.com 
>>>> <mailto:a11ythinker@gmail.com>> wrote:
>>>>> I’ve been building manually for now.
>>>>> *From:*Michael Cooper [mailto:cooper@w3.org]
>>>>> *Sent:*Tuesday, November 8, 2016 6:36 AM
>>>>> *To:*Michiel Bijl <michiel.list@moiety.me 
>>>>> <mailto:michiel.list@moiety.me>>; Matt King <a11ythinker@gmail.com 
>>>>> <mailto:a11ythinker@gmail.com>>
>>>>> *Cc:*ARIA Editors <public-aria-editors@w3.org 
>>>>> <mailto:public-aria-editors@w3.org>>
>>>>> *Subject:*Re: APG build fail
>>>>> I believe this is related to the sporadic non-functionality of 
>>>>> Respec that I've been seeing for weeks now. Shane has worked with 
>>>>> the systems people on this, and they did something to make it 
>>>>> happen less often, but it still happens. In particular, it seems 
>>>>> that to get respec-enabled documents to display properly in some 
>>>>> browsers it requires a refresh [Chrome] (in other browsers 
>>>>> [Firefox] even that doesn't work and it's just random luck as to 
>>>>> whether Respec works). Because the spec-generator service doesn't 
>>>>> refresh, it just fails. Shane said the spec-generator doesn't use 
>>>>> a browser, but it does seem to me to fail in the same frequency 
>>>>> that browsers do
>>>>> I still don't know what is causing the failure - a change in 
>>>>> browser security models, some subtle thing in the source of our 
>>>>> documents that isn't obvious, a change to W3C anti-denial of 
>>>>> service practices, some strange bug in Respec - so I haven't known 
>>>>> who to file a bug report with. Shane's attempt to work with the 
>>>>> W3C systems people seems to have produced some results but not 
>>>>> resolved the problem; I believe they're having a hard time 
>>>>> duplicating the problem even though I see it on nearly every 
>>>>> document I work with.
>>>>> I've been thinking of trying to come up with a workaround to this, 
>>>>> by including Respec as a library in our repositories rather than 
>>>>> pointing to its location on the W3C server, but until I get the 
>>>>> aria-common library working I won't know how to make that work. 
>>>>> I've also thought of trying to install the spec generator locally 
>>>>> within the build script rather than rely on the online version 
>>>>> (that fails a lot in the best of circumstances), but the 
>>>>> documentation is atrocious and I can't figure out how to get 
>>>>> started. It's all on my todo list, and perhaps I need to escalate 
>>>>> with more knowledgeable people.
>>>>> In the meantime, since the problem seems to be sporadic, you can 
>>>>> always try to restart the build. For practices, go 
>>>>> tohttps://travis-ci.org/w3c/aria-practices/and hit "Restart 
>>>>> build". I just tried this and it didn't work, but sometimes I have 
>>>>> seen it work.
>>>>> Michael
>>>>> On 08/11/2016 8:32 AM, Michiel Bijl wrote:
>>>>>> Hi,
>>>>>> Ever since pull request #147 
>>>>>> <https://github.com/w3c/aria-practices/issues/147> was merged 
>>>>>> into master the build is failing again. I’m not quite sure how to 
>>>>>> fix this, but could somebody look at it? If you know what it is 
>>>>>> I’d love to learn for next time this happens.
>>>>>> —Michiel
Received on Monday, 14 November 2016 17:56:54 UTC

This archive was generated by hypermail 2.3.1 : Monday, 14 November 2016 17:56:54 UTC