Hey Martin, - thanks a lot for the very detailed reply. I appreciate the great efforts your team has made, and I understand that it's hard to keep everything in check especially on a large scale. :)
I dug further into the issue, and eventually, it seems to be a problem in the communication between hs-tls and Google servers: https://github.com/vincenthz/hs-tls/issues/152 . So my temporary fix was switching to a third-party reverse proxy of the Google Fonts service. Indeed, further transient Internet issues may sometimes happen when the connection is bad, so migrating back to BiocStyle sounds like a good long-term solution. Thanks again, -Nan -- https://nanx.me On Wed, Sep 6, 2017, at 14:48, Martin Morgan wrote: > On 09/05/2017 10:35 AM, Nan Xiao wrote: > > Hey Kasper, > > > > - I really appreciate your help. Although it didn't help answer the > > question directly, I think it would still be necessary to clarify > > some facts: > > 1. I agree that using "non-standard fonts" could bring better aesthetics > > for a particular group of people -- maybe not everyone. But the > > "occasional" build problem should not be a reason why this > > personalization is discouraged. In fact, such build problem should > > not happen a lot (or, at all). I have been using such non-standard > > fonts for my 9 regularly updated R packages on CRAN since 2015, and I > > didn't encounter a single vignette build issue. > > 2. I've used BioCStyle since 2014, but eventually decided to switch to > > my own style years later because of the frequent build errors it > > caused back then. Maybe that's only my personal, limited experience > > (and I've seen the recent BioCStyle v2 got a lot of improvements), > > but this is the main reason why I trust the document style > > controllable by myself more than the other options. > > It's great that CRAN is so efficient and robust; I really don't know how > they do it. > > I'll re-iterate with Kasper that these customizations cost us > (Bioconductor core team) time -- maybe your package fails only once > every 1000 days, but there are a thousand developers so... > > Fixing BiocStyle once fixes it for all developers, while fixing a custom > issue once fixes it for just one developer. > > And users can choose to build vignettes, so they'll have to > trouble-shoot whatever issues are unique to your package. > > There really is value all around in keeping things simple. > > As for your particular issue, it seemed to fail on 3 of the 5 builds > that have reported yesterday, 2 of the five today. Have you noticed this > for several days in a row? > > I did ask about pandoc on the builders; they have > > malbec1/2 are at version 1.19.1 > tokay1/2 have 1.17.2 > veracruz1/2 have 1.19.2. > > of course we try to keep these in sync, but also do not go to > extraordinary lengths to install the same version across platforms. > These have not changed in the time frame that would cause your builds to > experience recent failures. > > This seems very much like a transient Internet access issue. Transient > access is one of the most difficult to debug, especially on our 'build' > servers where the loads experienced by the machines can vary a lot, and > where the vagaries of Internet connections can be catastrophic for a > build but almost unobserved by an individual investigating the problem. > > Martin > > > Best, > > -Nan > > > > -- > > https://nanx.me > > > > > > On Tue, Sep 5, 2017, at 09:16, Kasper Daniel Hansen wrote: > >> Build errors like these (related to processing of vignettes) happens > >> occasionally. I _strongly_ suggests not not using non-standard fonts > >> or advanced layout features. The small advantage of this (better > >> aesthetics according to the package writer, but not always according > >> to everyone who reads it) does not outweigh the disadvantage of > >> package failure and associated time burden on the package authors and > >> the build system / admins. I strongly suggests using BiocStyle; I > >> know that the layout doesn't agree with everyone, but (1) you can > >> expect that the build system will support this and (2) aesthetics is > >> highly personal anyway.> > >> Best, > >> Kasper > >> > >> On Mon, Sep 4, 2017 at 12:38 PM, Nan Xiao <m...@nanx.me> wrote: > >>> Dear BioC, > >>> > >>> - I see some very recent build errors for my packages (vignettes) > >>> which>> use Google Fonts: > >>> > >>> pandoc: Could not fetch > >>> https://fonts.googleapis.com/css?family=Alegreya+Sans:400,400i,700 > >>> TlsExceptionHostPort (HandshakeFailed Error_EOF) > >>> "fonts.googleapis.com">> 443 > >>> > >>> For example > >>> https://bioconductor.org/checkResults/release/bioc-LATEST/Rcpi/malbec2-buildsrc.html>> > >>> There were no such issues building these vignettes before, and > >>> I'm not>> even sure if this is a new issue caused by Google or pandoc. > >>> Could you>> please shine some light on this? > >>> > >>> Thank you, > >>> -Nan > >>> > >>> -- > >>> https://nanx.me > >>> > >>> _______________________________________________ > >>> Bioc-devel@r-project.org mailing list > >>> https://stat.ethz.ch/mailman/listinfo/bioc-devel > > > > [[alternative HTML version deleted]] > > > > _______________________________________________ > > Bioc-devel@r-project.org mailing list > > https://stat.ethz.ch/mailman/listinfo/bioc-devel > > > > > This email message may contain legally privileged and/or confidential > information. If you are not the intended recipient(s), or the employee > or agent responsible for the delivery of this message to the intended > recipient(s), you are hereby notified that any disclosure, copying, > distribution, or use of this email message is prohibited. If you have > received this message in error, please notify the sender immediately by > e-mail and delete this email message from your computer. Thank you. _______________________________________________ Bioc-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel