It looks like it’s another month since the last update. This long span seems to be becoming a trend, but I had good reason, being on vacation for much of this one. Since, as mentioned in the last update, the home Internet was down for a bit, that meant a bit of catch-up after both of these periods. Additionally, this was also the period in which the Mass Rebuild for Fedora 32 was run, and I took a little break for that. So this last period was a bit of an up-and-down as far as packaging work goes.

The Mass Rebuild found about 20 or so of my packages that failed to build. Some of these were false positives, and some of them are due to changes in GCC 10, notably Fortran handling of rank mismatches and the switch to -fno-common for C. For R-Rmpfr, a patch was already written upstream, which I simply had to backport. For R-deldir, I wrote a patch and sent it upstream, and it’s in the latest release.

One other thing I’m happy about is that I’ve finally been able to fix the build of R-data.table on non-x86 architectures. This has been broken for several upstream releases, and I’ve not been able to update in Fedora since 1.12.0. I wrote and proposed two patches to upstream to fix this:

• Improve fread for very small or very large fp numbers — This fixes conversion of strings to floating-point numbers on armv7hl and ppc64le by following a procedure similar to R’s builtin methods.
• Use consistent types with fwriteMainArgs.nrow — This fixes incorrect types passed to string formatting functions. Such an error is something common enough in C that the compiler grew warnings for it, but it only works on the standard library by default. By accident or de facto standards, the original code seemed to work everywhere but armv7hl.

There’s one more patch in Fedora that may or may not be sent upstream, but now the package can build on all architectures again, so we’re back on the latest version.

Updated packages⌗

New packages⌗

• golang-github-ssgelm-cookiejarparser — Parses a curl cookiejar file into a Go http.CookieJar
• R-broom — Convert Statistical Analysis Objects into Tidy Tibbles
• R-modelr — Modelling Functions that Work with the Pipe

Reviews⌗

• python-django-prometheus — Django middlewares to monitor your application with Prometheus.io
• git-secrets — Prevents committing secrets and credentials into git repos
• python-requests-futures — Asynchronous Python HTTP Requests
• python-aiounittest — Test asyncio code more easily