[Vm-dev] Ridiculous failures...

[Nicolas Cellier]

Hi all,
the legacy pharo builds in root opensmalltalk-vm repository are failing
randomly.

I suspect that the build should have failed since the upgrade of openSSL to
1.1
It failed in libssh2-1.17.0 because it was incompatible with this openSSL
version.
It did not fail immediately *thanks* to the wonderful travis cache features.

Apparently the official fork still depends on OpenSSL 1.0.2, which is
evidence that division of forces pay off!

I think I mostly fixed the builds by upgrading libssh2 to 1.9.0.
But the last builds got the most ridiculous failures I've ever seen.

> The job exceeded the maximum log length, and has been terminated.

Recompiling those libraries is ridiculous, does indeed spit ridiculous
amount of log (27500 lines), and might fail for many reasons including
timeout. Nobody cares, but it's also bad for the planet !

We know the solution for long: either build the libraries in their
respective projects, publish as github release, then download the binaries
(we can clone the repository as needed); or download pre-build binaries for
the distro (yeah, the CI bots build on a specific distro anyway, so at
least on linux and windows, the binaries are available unless we really
insist on using a really old version).

Many thanks to the Pharo team which has legated this mess, and did fix the
problem meanwhile in their own fork. A very appreciated spirit and
attitude...

By respect toward the pharo community, the root opensmalltalk-vm still
build the pharo binaries. But I really have the impression to clean-up
after the dog of someone else. Respect has to be mutual, or it's not
respect, it's slavery.

Due to this lack of support, I had to put all the pharo builds in the list
of authorized failures in .travis.yml.

I kindly ask interested parties to help implement an equivalent solution in
the root repository, if those builds still have a value. Without community
support, you know that those bits will inevitably rot (the more the
dependencies, the faster the rot).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/vm-dev/attachments/20200617/15730596/attachment.html>