[Libwebsockets] issue tracking post github
andy at warmcat.com
Thu Jun 28 11:21:11 CEST 2018
On 06/28/2018 04:24 PM, Alfred Sawaya wrote:
> For what it worth, I agree with you at 100%. Mailing lists are a good
> What github did (and it still does) is standardizing the process to
> contribute and make it very simple and centralized (fork, implement,
> request pull, test pipelines).
> Back to old school then, mailing list, send patches (or rp but you would
> have a lot of remotes to handle)...
Well, I generally take your point.
But on lws github currently, there are 958 "issues" (955 closed) and 360
PRs (359 closed). So the main thing github generates are "issues".
It's complicated a bit because some issues generate patches, usually by
me, since when the issue describes something needing doing the raw
material for these patches is sometimes showing changed code but most
often very abstract and not a diff or whatever, but prose.
On the other side, some of the "issues" are just lazy garbage.
If people want to push patches on github and point me to them because
that's easier for them, I'm not going to ignore their contribution. But
github will just become a mirror going on when my side is straight.
This is the approach taken by, eg git itself
> Alfred Sawaya
> Message original
> De: andy at warmcat.com
> Envoyé: 28 juin 2018 2:18 AM
> À: karlp at tweak.net.au; libwebsockets at ml.libwebsockets.org
> Objet: Re: [Libwebsockets] issue tracking post github
> On 06/27/2018 08:14 PM, Karl Palsson wrote:
> > I see that you're moving heavily away from github.
> > What are you plans for issue tracking in this new world?
> Currently my best plan is this mailing list.
> I have written a load of patches on cgit that improve it towards the
> relevant features from github. For example display the top level README
> in the directory view with markup, ability to show highlight ranges etc
> But getting them upstream is slow, that project has been around for > 20
> years, most of the devs there are sniffy about the idea of introducing
> the patches.
> If it turns out to be impossible to get these features in cgit I will
> have to change to a plan B.
> Well that's the answer, but here is a more in-depth musing --->
> There's a larger philosophical question about "issues"... what is it for?
> "Obviously" it's for helping users use the library. But from my
> perspective that's not really true... it's for learning about problems
> in lws and getting contribution for mutual benefit. Many times I saw
> some question there about integration of lws in proprietary software
> from someone who will never contribute anything back and asked myself
> "why do I sit here and answer these". Actually, there is no motive at
> all for that case. Nothing for me improves if the number of such users
> doubles or triples, just more "issues" correspondingly.
> These type of issues became overwhelming a few months ago and triggered
> (it had already been suggested but I refused it, since it's a lot of
> work to do and maintain) the creation of the minimal-examples to stop
> the same questions over and over. But that too was actually a painful
> form of "learning about problems in lws".
> Other times I get awesome issues where the guy found real problems, and
> points me to them clearly (sometimes a textual explanation, or some line
> numbers or function names is enough)... other times fix patches come
> from nowhere... discussion of larger but real issues... these are all
> great and help me and everyone using lws.
> Another problem with "issues" is that it just apes proprietary
> commercial support flow, and user expectations. But this isn't
> proprietary commercial software and as far as users are potential
> contributors, they are not some baby bird screeching to be fed and owed
> a response, but simply people in the same boat as the developer. We all
> want lws to work well and maintainably. Sometimes debugging into lws is
> hard if you don't know what you're looking at, but most times, nothing
> stops users digging a bit and characterizing / confirming the origin of
> the problem themselves before raising it. I think mailing lists tend to
> encourage that more than "issues".
> > Sincerely,
> > Karl Palsson
> > _______________________________________________
> > Libwebsockets mailing list
> > Libwebsockets at ml.libwebsockets.org
> > https://libwebsockets.org/mailman/listinfo/libwebsockets
> Libwebsockets mailing list
> Libwebsockets at ml.libwebsockets.org
More information about the Libwebsockets