[tex-live] Stable, reproducible, scriptable TexLive installation

Alistair Miles alimanfoo at googlemail.com
Fri Aug 18 00:24:00 CEST 2017


Hi Karl,

On Thu, Aug 17, 2017 at 6:31 PM, Karl Berry <karl at freefriends.org> wrote:

>     recently the server seems to have become flaky
>
> Nothing has changed wrt tug.org. I surmise any "flakiness" is because
> there is (and has always been) a noticeable limit on ftp
> connections. I've done that because in my experience web browsers tend
> to use up all available ftp connections anyway, and it was untenable.
>
>     be grateful if anyone has any other suggestions.
>
> Use rsync://tug.org/historic/... instead of ftp.
> Or rsync://rsync.math.utah.edu/historic/... for Nelson's mirror.
>

Thanks, will give it a try.


> But also, I don't understand you're doing. You're installing TL16 But
> TL16 is never going to change again. So where does "continuous
> integration" come in? And "continuous integration" and "stable" are
> polar opposites anyway.  I don't get it. And why not use TL17? Oh well,
> whatever.
>

I'm writing a paper with latex source in a github repo. Whenever anyone (me
or collaborator) makes a PR to incorporate some changes, the CI service (I
use travis) automatically builds the PDF. So the CI service is a way to
check if someone has made changes that break the build, and to avoid any
breaking changes getting into the master branch.

I'm using TL16-final to guarantee that texlive won't change under our feet.
I.e., if I run my texlive install script today, it will have the same
results as running it in a week or a month. This is very helpful to get all
collaborators using exactly the same texlive software versions, and also
getting exactly the same installation on the CI server.

This is similar to pinning dependencies when working on a software or
analysis project. E.g., we use conda to install Python for our data
analysis work, and we pin versions of all installed Python packages so all
team members are using exactly the same software, and everything is
consistently reproducible.


>     historic/ was mirrored along with all the other texlive content.
>
> It's not feasible to dump XX more gigabytes of old unchanging stuff onto
> ctan.org and every ctan mirror. CTAN does not hold old versions of
> anything, in general.
>

Fair enough if space is a premium. Just wanted to register that for us,
being able to get an exactly reproducible texlive install is very useful.

Cheers,
Alistair


>
> Happy TeXing,
> Karl
>



-- 
Alistair Miles
Head of Epidemiological Informatics
Centre for Genomics and Global Health <http://cggh.org>
Big Data Institute Building
Old Road Campus
Roosevelt Drive
Oxford
OX3 7LF
United Kingdom
Phone: +44 (0)1865 743596
Email: alimanfoo at googlemail.com
Web: http://a <http://purl.org/net/aliman>limanfoo.github.io/
Twitter: https://twitter.com/alimanfoo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://tug.org/pipermail/tex-live/attachments/20170817/071c82ed/attachment-0001.html>


More information about the tex-live mailing list