Hi Ed As you might gather from the delayed reply to this email, I have limited time for this. That includes things like ‘learning how this CI thing works’. But I have automated this to some extent and hope to do more. It would be nice to have the GitHub CI take care of it, if such a thing is possible?
Right now I would like some people to try out the docker container, and report back. Karl > On 17 Aug 2021, at 11:58 am, Ed . <ej...@hotmail.com> wrote: > > Hi Karl, > > In the Agile methodology, they say if e.g. making a release is painful, do it > more so that you’ll be forced to automate away the pain. I’d suggest a > similar benefit would arise if you released SciPDL more often, but it’s up to > you. > > Having thought more about this, I think the existing PDLPorters/devops repo > would be ideal: could you please push on a branch of that, making a new > directory within it called something like build-docker, putting your files in > there, then make a PR so that people can take a look? > > Regarding the number of releases of PDL, my approach has been to release it > when I thought something valuable was ready. You’ll recall that in the past, > just because PDL wasn’t released very often, didn’t mean every single release > was perfect. There’s one bit in a release notes from the past proudly noting > a release with over 50 fixed tickets. Those fixes were sat on for a > considerable period of time, with users unable to benefit from them. > > We do have literal “continuous integration” (using GitHub Actions) which has > all the tests Zaki and I could think of to throw at it, for every single push > on any branch: it builds on Ubuntu, CentOS, MacOS, Windows, Cygwin (only for > releases, it’s too slow otherwise). It would be easy(ish) to have, probably > only for releases, an additional build step that updated a Docker image with > SciPDL (and as an added benefit, would show any breakages of that). The CI > reports all pushes, and build results, on the IRC channel, which I do > recommend sitting on. Having SciPDL be an additional “canary in the coal > mine” would be of considerable value. > > As for which is the last “stable release”, I understand the question, but for > the reasons in the last two paragraphs I can’t really accept the premise. If > you look at https://metacpan.org/pod/PDL <https://metacpan.org/pod/PDL>, > there’s a not-very-up-to-date “Testers” readout (the numbers are cached, it > only shows currently 100 passes), and link. If you follow the link, the > current version shown is for 2.057 (the latest), > http://matrix.cpantesters.org/?dist=PDL+2.057 > <http://matrix.cpantesters.org/?dist=PDL+2.057>: it shows 131 passes, 7 > unknown (which are build failures, each of which I know the reason for), and > no failures (either 100% passes, or 95%, depending on how you count it), > across 70 configurations (OS, Perl version). How “stable” were you after, > exactly? 😉 > > Best regards, > Ed > > From: Karl Glazebrook <mailto:karlglazebr...@mac.com> > Sent: 17 August 2021 01:56 > To: Ed . <mailto:ej...@hotmail.com> > Cc: perldl <mailto:pdl-general@lists.sourceforge.net>; Bob Abraham > <mailto:abra...@astro.utoronto.ca> > Subject: Re: [Pdl-general] SciPDL Docker > > Hi Ed, > > It’s a Dockerfile and a build shell script that runs inside it. I went that > way as that is what I did for MacOS and docker seemed a nice way of bypassing > all the issues with Debian packaging (see other thread) which are frankly > doing my head in! > > I am not sure I feel comfortable sharing my dubious build scripts, could be > dangerous, but maybe if it sits within PDL repo. > > BTW I only make a new SciPDL once or twice a year. It’s kind of my thing to > bundle up all the stuff I normally like but others find it useful. So - I see > there have been a huge number of new PDL versions this year, which is > fantastic (esp. the new complex numbers approach) but I also see things are > in a high state of flux. Continuous integration of new changes are not my > thing, at least not for SciPDL, what would be the last ’stable version’ to > build against do you think? I hope that question makes sense, > > Karl > > > > On 17 Aug 2021, at 1:01 am, Ed . <ej...@hotmail.com > <mailto:ej...@hotmail.com>> wrote: > > Hi Karl, > > That’s great! Can you share your Dockerfile? Is it on GitHub? I’m thinking it > would be great to have it within PDLPorters, maybe in a repo called (very > imaginatively) “docker”. > > Yes, there is in fact now a 2.057 (which restored the DELETEDATA mechanism > which it turns out people were using for other than mmap – oops). Please give > it a go! > > Best regards, > Ed > > From: Karl Glazebrook via pdl-general > <mailto:pdl-general@lists.sourceforge.net> > Sent: 16 August 2021 13:11 > To: perldl <mailto:pdl-general@lists.sourceforge.net> > Cc: Bob Abraham <mailto:abra...@astro.utoronto.ca> > Subject: [Pdl-general] SciPDL Docker > > Hi PDL users, > > I made a Docker version of SciPDL and put it on Dockerhub. It was on my to-do > list for a while, helped me learn more about Docker. > > So you can run it anywhere you can run Docker with a command like: > > docker run -it karlglazebrook/scipdl pdl > > It has pgplot (make sure to set X11 DISPLAY for this) and all the usual > stuff etc. I include in my SciPDL ‘kitchen sink' > > This is still on PDL-2.025, same as MacOS, and is an intel builf, next round > to it is to update the PDL versions in SciPDL. (Did I see v56 recently! > Jeepers...) > > best > > Karl
_______________________________________________ pdl-general mailing list pdl-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/pdl-general
