[Bioc-devel] RMassBank build error

[Martin Morgan]

On 10/07/2016 01:52 PM, Laurent Gatto wrote:
>
> Dear Emma,
>
> The error is very strange indeed, and I hope Martin can help us out here.

I did look at this a little but not completely, and it seems like you've 
come up with another solution? So I'll stop.

also for the intermittent error you mention, I skimmed the issue and saw 
the problem with the reference class finalizer. Roughly my understanding 
of what happens is

env = new.env()
env$f = function(...) message("done")
reg.finalizer(env, env$f)
rm(env); gc()

The garbage collector runs, marking env and other objects as available 
for re-use. Once the garbage collector is finished (not just collecting 
env, but all objects), any finalizers that are now eligible for 
evaluation are run. So there is a chance that the environment itself has 
been re-used by the time the finalizer runs, leading to intermittent errors.

This makes finalizers on reference classes unreliable.

Martin

>
> With the latest RMassBank and MSnbase, I get
>
>> new("RmbSpectrum2")
> Error in initialize(value, ...) :
>   'initialize' method returned an object of class “Spectrum2” instead of
>   the required class “RmbSpectrum2”
>
> which reproduces the error.
>
> The RmbSpectrum2 class is defined in a standard way
>
> .RmbSpectrum2 <- setClass("RmbSpectrum2",
>                           representation = representation(
>                               satellite="logical",
>                               low="logical",
>                               rawOK ="logical",
>                               good = "logical",
>                               mzCalc = "numeric",
>                               formula = "character",
>                               dbe = "numeric",
>                               formulaCount = "integer",
>                               dppm = "numeric",
>                               dppmBest = "numeric",
>                               ok = "logical",
>                               info = "list"
>                           ),
>                           contains=c("Spectrum2"),
>                           prototype = prototype(
>                               satellite = logical(),
>                               low = logical(),
>                               rawOK = logical(),
>                               good = logical(),
>                               mzCalc = numeric(),
>                               formula = character(),
>                               dbe = numeric(),
>                               formulaCount = integer(),
>                               dppm = numeric(),
>                               dppmBest = numeric(),
>                               ok = logical(),
>                               info = list(),
>                               new("Versioned", versions=c(classVersion("Spectrum2"),
>                                                           RmbSpectrum2 = "0.1.0"))
>                           ))
>
>
> and
>
>> new("Spectrum2")
> Object of class "Spectrum2"
>  Precursor: NA
>  Retention time: :
>  Charge: NA
>  MSn level: 2
>  Peaks count: 0
>  Total ion count: 0
>
> works.
>
> The MSnbase maintainers have had a bit of a struggle with spurious and
> stange failures in the recent past (see [1]). This ans a whole new
> backend in the package have led to the following initialize method, that
> constructs the class directly in C
>
> setMethod("initialize",
>           "Spectrum2",
>           function(.Object, msLevel = 2L, peaksCount = length(mz),
>                    rt = numeric(), acquisitionNum = NA_integer_,
>                    scanIndex = integer(), tic = 0L, mz = numeric(),
>                    intensity = numeric(), fromFile = numeric(),
>                    centroided = NA, smoothed = NA,
>                    polarity = NA_integer_, merged = 1,
>                    precScanNum = NA_integer_, precursorMz = NA,
>                    precursorIntensity = NA, precursorCharge = NA_integer_,
>                    collisionEnergy = NA) {
>               .Object <- Spectrum2_mz_sorted(msLevel, peaksCount, rt,
>                                              acquisitionNum, scanIndex,
>                                              tic, mz, intensity, fromFile,
>                                              centroided, smoothed, polarity,
>                                              merged, precScanNum, precursorMz,
>                                              precursorIntensity, precursorCharge,
>                                              collisionEnergy)
>               if (validObject(.Object))
>                   .Object
>           })
>
>
> Why is calling new("RmbSpectrum2") direclty returning an Spectrum2
> object? Is this due to us not calling callNextMethod?
>
> Laurent
>
> [1] https://github.com/lgatto/MSnbase/issues/138
>
>
> On  7 October 2016 17:02, Schymanski, Emma wrote:
>
>> Hi BioC team,
>>
>> In all the mzR troubles, it slipped through that RMassBank had a build error of it's own presumably caused by an update to MSnbase - for some reason we never received the email with the build error reports?!
>> We have discovered the error now, very late, and are onto finding out the cause to fix it but it is not straightforward. Now that it is the day of the deadline to pass build without error, are we able to have a little leeway if needed? It's taken us the whole day to get the right binaries to actually have a chance to start fixing...
>> Can someone also check or explain why we no longer receive the emails reporting errors to us?
>>
>> Thanks,
>> Emma (on behalf of the others)
>> _______________________________________________
>> Bioc-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/bioc-devel
>
>


This email message may contain legally privileged and/or...{{dropped:2}}