[Bioc-devel] FYI: Windows build of affyPLM devel has been broken for a while

[Henrik Bengtsson]

Hi.

On 9/22/07, Martin Morgan <mtmorgan at fhcrc.org> wrote:
> "Henrik Bengtsson" <hb at stat.berkeley.edu> writes:
>
> > Hi,
> >
> > I noticed that someone (you) fixed it recently.  Thanks.  Yes,
> > spawning ghostscript in a parallel thread/process in the background is
> > not easy to control.  I've once tried to control for this by probing
> > the output file for existence and file size and when the file size
> > wasn't changing for a certain time, I let the code assume that it was
> > ready.  I cannot remember if also tried to have a dummy file generated
> > afterwards that could check for.  It was a while ago, but there are
> > some hacks you can do beyond Sys.sleep().
>
> Tom Short (responding off-list to my query on R-devel) also mentioned
> hacking bitmap to call ghostscript directly, and using Cairo / GDD
> (though Tom did not have direct experience in vignettes) which are
> faster than ghostscript. A solution in R would warrant more thorough
> approaches.
>
> > However, in the bigger picture, have you/it been discussed to have
> > failed vignettes not to generate an error/stop the build of the
> > package, and instead let R CMD check give the error?  Is that even
> > possible?  A big tryCatch() around the vignette compiler?  In some
>
> Not sure what the official position is on this, but vignettes serve,
> for better or worse, as a kind of integration test. Failed vignettes
> usually turn up meaningful errors. Of course a better practice would
> exploit formal test structures.

Yes, I am aware of that, but all other test cases are validated by R
CMD check not R CMD build.

I believe allowing vignettes to fail during build comes down to how
vignette is important for the service the package provides.  For
instance, I can definitely see how code and data (possibly not example
data but is currently no such separation) should be valid in order to
install a package.  Documentation, however, is not critical for a
package's operation.  To me a vignette is just another documentation
format just like the TeX, HTML, CHTML, man pages, ...

Don't get me wrong, I do believe packages should build and pass all
builds and checks without errors and warnings.

Cheers

Henrik


>
> Martin
>
> > sense a broken vignette is like a broken example in a Rd file - it is
> > not critical for the package, but for the documentation, which is
> > something for R CMD check to "complain about".
> >
> > Cheers
> >
> > Henrik
> >
> >
> >
> > On 9/22/07, Martin Morgan <mtmorgan at fhcrc.org> wrote:
> >> Hi Henrik --
> >>
> >> The problem appears to be that 'bitmaap / dev.off' on windows spawns
> >> ghostscript to perform the implied graphics file conversion. R does
> >> not wait for the process to return, and Sweave cleans up before
> >> ghostscript completes.
> >>
> >> The more-or-less ugly solution is to insert a Sys.sleep(10) in the
> >> vignette. Whether to insert this, and the duration of sleep, entirely
> >> depends on the build system and vignette. For instance, MAplots.Rnw
> >> uses bitmap and needs to sleep, but QualityAssess.Rnw uses bitmap but
> >> does not need to sleep.
> >>
> >> Martin
> >>
> >> "Henrik Bengtsson" <hb at stat.berkeley.edu> writes:
> >>
> >> > Hi,
> >> >
> >> > the affyPLM devel build for Windows has been broken for a while.  Last
> >> > successful build was/is v1.13.6 (2007-06-22) and now were on
> >> > v1.13.13(14).  Since there has recently been quite of bit of migration
> >> > from affyPLM to preprocessCore, code that relies on
> >> > affyPLM/preprocessCore is out of date on Windows.
> >> >
> >> > I'm not sure this is a package maintainer problem or a problem with
> >> > the BioC builds. From the build report:
> >> > http://bioconductor.org/checkResults/2.1/bioc-LATEST/affyPLM/liverpool-buildsrc.html
> >> >
> >> > texi2dvi.exe: Permission denied: MAplots.tex
> >> > Error in texi2dvi(file = bft, pdf = TRUE, clean = FALSE, quiet = quiet) :
> >> >   running 'texi2dvi' on 'MAplots.tex' failed
> >> > Calls: buildVignettes -> texi2dvi
> >> > Execution halted
> >> > Error: /undefinedfilename in
> >> > (E:\\biocbld\\bbs-2.1-bioc\\tmpdir\\RtmphbJVrR\\Rbit35182069)
> >> > Operand stack:
> >> >
> >> > Execution stack:
> >> >    %interp_exit   .runexec2   --nostringval--   --nostringval--
> >> > --nostringval--   2   %stopped_push   --nostringval--
> >> > --nostringval--   --nostringval--   false   1   %stopped_push
> >> > Dictionary stack:
> >> >    --dict:1126/1686(ro)(G)--   --dict:0/20(G)--   --dict:70/200(L)--
> >> > Current allocation mode is local
> >> > Last OS error: No such file or directory
> >> > AFPL Ghostscript 8.54: Unrecoverable error, exit code 1
> >> >
> >> > Cheers
> >> >
> >> > Henrik
> >> >
> >> > _______________________________________________
> >> > Bioc-devel at stat.math.ethz.ch mailing list
> >> > https://stat.ethz.ch/mailman/listinfo/bioc-devel
> >>
> >> --
> >> Martin Morgan
> >> Bioconductor / Computational Biology
> >> http://bioconductor.org
> >>
>
> --
> Martin Morgan
> Bioconductor / Computational Biology
> http://bioconductor.org
>