Re: [Bioc-devel] RMassBank (almost) builds ...
Hi, Is there any compelling reason why one wouldn't want to upgrade BioC to Java 8? One way or another this update will have to come in the near future anyway, so causing disruption by downgrading rcdk now, then upgrading it again in the future seems unnecessary... From: Schymanski, Emma Sent: Donnerstag, 13. Oktober 2016 16:39 To: Stravs, Michael Subject: FW: [Bioc-devel] RMassBank (almost) builds ... Sent: Thursday, 13 October 2016 4:33 PM To: Rajarshi Guha mailto:rajarshi.g...@gmail.com>> Cc: Egon Willighagen mailto:egon.willigha...@gmail.com>>; Schymanski, Emma mailto:emma.schyman...@eawag.ch>>; bioc-devel@r-project.org<mailto:bioc-devel@r-project.org> Subject: Re: [Bioc-devel] RMassBank (almost) builds ... Hi Rajarshi, It was my understanding that the newer depict module needed Java8 and that was the motivating reason for setting the system requirement to Java8. For example, when i was drawing molecules on my local machine I would encounter errors using Java7 and CDK >=1.5.12 , but did not encounter these with Java8. If i remember correctly the basic image generation test was failing with Java7 (https://github.com/cran/rcdk/blob/master/inst/unitTests/runit.depiction.R). When I initially submitted to CRAN I did NOT check the Java version and therefore users would be able to use Java 7 and would encounter errors only when using some features of Depict. But in going through the submission gauntlet I was asked to add a system requirement and a check for the Java version. In retrospect a warning may have been more appropriate. In terms of fixing this issue, I would mention that the single biggest problem in uploading to CRAN is that the CDK JAR size is much bigger (~18mb?) than the maximum allowed size (~5mb). I was given a waiver when I argued that the CDK core libraries are updated only infrequently. I think it would be very nice to have more rapid updates of rCDK on CRAN that tracks the CDK releases but this would require a better way to load JARs on CRAN as the package size was a friction point. For comparison, many rJava packages use only a few Java files that boil down to small Jars of <1Mb. The admins were understandably concerned about the strain on the archive system of large JARs. That was a bit of an aside but I bring this up to say that even a small change like downgrading a Java version check to a warning or changing the SystemRequirements line will require a full upload of the 18MB jar which presented an issue upon last submission. In my view the potential solutions can be: 1. if RMassBank doesn't need the depict stuff, pin the rCDK version to 3.3.5 (CDK libs 1.5.12 but no Java check) 2. update BioC buildsystem (as per Herve's email and keep J8) 3. change the System requirements to java7 and re-upload to CRAN (are we sure? Egon would know but my own experience suggests there are issues here.) 4. Long-term: is there a better way to package JARs for CRAN that mitigate the package-size issue and the resulting reluctance to make small, incremental changes? Perhaps a Maven plugin? zach cp On Thu, Oct 13, 2016 at 8:46 AM, Rajarshi Guha mailto:rajarshi.g...@gmail.com>> wrote: Hi Zach, I see from the commits that you've put in a check for JDK >= 1.8 in the init methods and the unit tests. Is there a reason for that? According to Egon, CDK currently depends on >= 1.7 On Thu, Oct 13, 2016 at 4:15 AM, Egon Willighagen mailto:egon.willigha...@gmail.com>> wrote: cc: Rajarshi Hi Emma, On Wed, Oct 12, 2016 at 10:18 PM, Schymanski, Emma mailto:emma.schyman...@eawag.ch>> wrote: > Thanks to Laurent and Jo for updating MSnbase and thanks to the BioC team for > getting the email updates working again! The good news is that RMassBank now > builds on most platforms... but ... > 1) we have a build error due to a Java version issue with rcdk, this has been > a problem on release for ages. > http://bioconductor.org/checkResults/3.4/bioc-LATEST/RMassBank/morelia-buildsrc.html > Will someone on the BioC side follow up with that? It seems that Rajarshi's package depends on J8 which is not present on the build system? I did a quick search, but is rcdk actually in Bioconductor? I guess it is using the CRAN version, which has this dependency formalized: SystemRequirements:Java JDK 1.8 or higher So, the RMassBank is effectively also depending on J8... is that available on the build system? Rajarshi, the CDK depends only on J7... why is the rcdk depending on J8? Egon -- E.L. Willighagen Department of Bioinformatics - BiGCaT Maastricht University (http://www.bigcat.unimaas.nl/) Homepage: http://egonw.github.com/ LinkedIn: http://se.linkedin.com/in/egonw Blog: http://chem-bla-ics.blogspot.com/ PubList: http://www.citeulike.org/user/egonw/tag/papers ORCID: -0001-7542-0286 ImpactStory: https://impactstory.org/u/egonwillighagen -- Rajarshi Guha
Re: [Bioc-devel] RMassBank (almost) builds ...
Hi all, including all BioC developers, On Thu, Oct 13, 2016 at 4:32 PM, Zach Charlop-Powers wrote: > When I initially submitted to CRAN I did NOT check the Java version and > therefore users would be able to use Java 7 and would encounter errors only > when using some features of Depict. But in going through the submission > gauntlet I was asked to add a system requirement and a check for the Java > version. In retrospect a warning may have been more appropriate. John May told me that it only depends on J7, but maybe one of the CDK dependencies now depends on J8? > In terms of fixing this issue, I would mention that the single biggest > problem in uploading to CRAN is that the CDK JAR size is much bigger > (~18mb?) No, more like 25MB :) > than the maximum allowed size (~5mb). Ah... well, it's mostly dependencies that increase the size. BioC developers, would rcdk be welcome on BioConductor? It's frequently used in biology, e.g. metabolomics. Rajarshi, Zach, would you consider that if the size issue becomes an issue? Alternative, the whole CDK modularization has the purpose of being able to reduce the size of the needed libs. We could look and see which CDK modules are actually used by rcdk and then not depend on the full CDK but only on the modules it really needs. > I was given a waiver when I argued that the CDK core libraries are updated > only infrequently. That only makes sense of you put the CDK dependencies in rcdk-libs, or possible split up in three versions, because the 3rd party libs the CDK depends on are more stable than the CDK, which has releases every half year or so at this moment. I think if you put the 3rd party lib jars in rcdk-libs and the CDK jars in rcdk itself, you may still be below the 5MB... > I think it > would be very nice to have more rapid updates of rCDK on CRAN that tracks > the CDK releases but this would require a better way to load JARs on CRAN as > the package size was a friction point. For comparison, many rJava packages > use only a few Java files that boil down to small Jars of <1Mb. The admins > were understandably concerned about the strain on the archive system of > large JARs. Yes, I know this issue, which I had with the rrdf package too. (BioC developers, would that package be welcome on BioConductor?) Egon -- E.L. Willighagen Department of Bioinformatics - BiGCaT Maastricht University (http://www.bigcat.unimaas.nl/) Homepage: http://egonw.github.com/ LinkedIn: http://se.linkedin.com/in/egonw Blog: http://chem-bla-ics.blogspot.com/ PubList: http://www.citeulike.org/user/egonw/tag/papers ORCID: -0001-7542-0286 ImpactStory: https://impactstory.org/u/egonwillighagen ___ Bioc-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
Re: [Bioc-devel] RMassBank (almost) builds ...
Hi Emma, On 10/12/2016 01:18 PM, Schymanski, Emma wrote: Hi all, Thanks to Laurent and Jo for updating MSnbase and thanks to the BioC team for getting the email updates working again! The good news is that RMassBank now builds on most platforms... but ... 1) we have a build error due to a Java version issue with rcdk, this has been a problem on release for ages. http://bioconductor.org/checkResults/3.4/bioc-LATEST/RMassBank/morelia-buildsrc.html Will someone on the BioC side follow up with that? We're aware of this and will try to install Java 1.8 on morelia soon. It might not happen in time for the release though, sorry. Please ignore the error on morelia for now. 2) Our build error notification said the problem was with malbec1 http://bioconductor.org/checkResults/3.4/bioc-LATEST/RMassBank/malbec1-buildsrc.html We seem to be OK and OK but "build bin" is blank - is this a problem? It's blank because we don't build package binaries on Linux. Only on Windows and Mac. If so, how can we find the error message, because it does not appear in the build report for malbec1? Is it also related to java? The fact that you received a build failure notification with a link to a result that is actually OK is probably due to the quirk I explained in my email from yesterday. See: https://stat.ethz.ch/pipermail/bioc-devel/2016-October/009882.html I think I've addressed the issue with the script so hopefully this shouldn't happen again today. So today you shouldn't receive any build failure notification because we only send them for failures on the Linux builder (i.e. malbec1). Please let me know if this happens again. Thanks, H. Thanks, Emma [[alternative HTML version deleted]] ___ Bioc-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel -- Hervé Pagès Program in Computational Biology Division of Public Health Sciences Fred Hutchinson Cancer Research Center 1100 Fairview Ave. N, M1-B514 P.O. Box 19024 Seattle, WA 98109-1024 E-mail: hpa...@fredhutch.org Phone: (206) 667-5791 Fax:(206) 667-1319 ___ Bioc-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
Re: [Bioc-devel] RMassBank (almost) builds ...
cc: Rajarshi Hi Emma, On Wed, Oct 12, 2016 at 10:18 PM, Schymanski, Emma wrote: > Thanks to Laurent and Jo for updating MSnbase and thanks to the BioC team for > getting the email updates working again! The good news is that RMassBank now > builds on most platforms... but ... > 1) we have a build error due to a Java version issue with rcdk, this has been > a problem on release for ages. > http://bioconductor.org/checkResults/3.4/bioc-LATEST/RMassBank/morelia-buildsrc.html > Will someone on the BioC side follow up with that? It seems that Rajarshi's package depends on J8 which is not present on the build system? I did a quick search, but is rcdk actually in Bioconductor? I guess it is using the CRAN version, which has this dependency formalized: SystemRequirements:Java JDK 1.8 or higher So, the RMassBank is effectively also depending on J8... is that available on the build system? Rajarshi, the CDK depends only on J7... why is the rcdk depending on J8? Egon -- E.L. Willighagen Department of Bioinformatics - BiGCaT Maastricht University (http://www.bigcat.unimaas.nl/) Homepage: http://egonw.github.com/ LinkedIn: http://se.linkedin.com/in/egonw Blog: http://chem-bla-ics.blogspot.com/ PubList: http://www.citeulike.org/user/egonw/tag/papers ORCID: -0001-7542-0286 ImpactStory: https://impactstory.org/u/egonwillighagen ___ Bioc-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel