diff: "Error in grid.newpage() : internal read error in PDF_endpage"

[funnell]

Hello,

I'm getting an error when trying to run vast-tools diff:

/share/lustre/tfunnell/local/software/vast-tools/vast-tools diff -a SA464_A14559_D0R1PACXX_8_1 -b SA465_A14560_D0R1PACXX_8_1 -c 4 -o /share/scratch/tfunnell_temp/vast/output > /share/scratch/tfunnell_temp/vast/output/SA464_A14559_D0R1PACXX_8_1_VS_SA465_A14560_D0R1PACXX_8_1_INCLUSION-FILTERED.tab
Trying to load required package: optparse
Trying to load required package: RColorBrewer
Trying to load required package: reshape2
Trying to load required package: ggplot2
Trying to load required package: grid
Trying to load required package: parallel
Error in grid.newpage() : internal read error in PDF_endpage
Calls: plotPrint -> grid.newpage
Execution halted

Running the following in R 3.0.2 doesn't cause any errors:

> library(grid)
> grid.newpage()

[timbitz]

Hi, I don't think I've seen that error before, so that is odd.

Can you try that command without the -c flag. And also can you add pdf("test") before the grid.newpage in your test example...

Are you on the most recent commit?

Thanks!

[timbitz]

Hi @funnell,

As a follow up, I think we can rule out R version discrepancy. I updated to the latest version of R, 3.1.1 (from 3.0.1), and using the latest commit of vast-tools be04038, diff runs successfully and produces proper output (filtered and pdf) on real data. So something else might be going on here.

bc2: ./install.R 
Using R version 3.1.1 (2014-07-10)
Looking for VAST Database [VASTDB]
Found what appears to be VASTDB.. OK
Trying to load required package: getopt
Trying to load required package: optparse
Trying to load required package: RColorBrewer
Trying to load required package: reshape2
Trying to load required package: ggplot2
Trying to load required package: grid
Trying to load required package: parallel
Setting vast-tools permissions... success!
/home/blencowe/blencowe2/timsw/bin/bowtie-1.0.0/bowtie
Found bowtie... Everything looks --OK

Then:

bc2: vast-tools diff -a SampleA -b SampleB -o vast -i INCLUSION_LEVELS_FULL-Hsa2.tab -c 4
Trying to load required package: optparse
Trying to load required package: RColorBrewer
Trying to load required package: reshape2
Trying to load required package: ggplot2
Trying to load required package: grid
Trying to load required package: parallel
GENE    EVENT   COORD   LENGTH  FullCO  COMPLEX SampleA SampleA-Q   SampleB SampleB-Q
MYO9A   HsaEX0041285    chr15:72311380-72311436 57  chr15:72313259,72311380-72311436,72302788   S   77.78   OK,OK,OK,OK,S@49.78,14.22   43.33   LOW,LOW,LOW,OK,S@18.63,24.37
TDRD3   HsaEX0064323    chr13:61068569-61068709 141 chr13:61060082,61068569-61068709,61083897   S   100.00  OK,OK,OK,OK,S@56.00,0.00    78.38   OK,OK,OK,OK,S@51.73,14.27
...

And panels from the output:

[funnell]

Hi,

Removing the -c parameter seemed to fix it. This might have something to do with me running vast-tools on a shared filesystem... it's caused me problems in the past.

[timbitz]

Are you up to date with the current commit? It was not so long ago that I fixed a few bugs relating to plotting and multicore function... There are a few weird things that mclapply seems to do with graphics open, but none of the forks write to the graphics device, just to a list, which is subsequently printed to graphics post processing... So I'm not sure how or whether the filesystem could affect things here. But for sure if you have a commit from before that fix, 438eca5 ... then that might explain it.

[funnell]

Yup, I'm up to date. I did some more testing though and found that the error doesn't happen every time I run the diff command. I ran

/share/lustre/tfunnell/local/software/vast-tools/vast-tools diff -a SA464_A14559_D0R1PACXX_8_1 -b SA465_A14560_D0R1PACXX_8_1 -c 6 -o /share/scratch/tfunnell_temp/vast/output > /share/scratch/tfunnell_temp/vast/output/SA464_A14559_D0R1PACXX_8_1_VS_SA465_A14560_D0R1PACXX_8_1_INCLUSION-FILTERED.tab

which failed, then I ran it again immediately afterwards and it finished without error.

[timbitz]

Thanks for this. Yah thats sounds like something counterintuitive is happening behind the scenes with the forks during the var <- ggplot() calls within mclapply I think. I will look into it.

[timbitz]

It looks like that error might have something to do with pdf compression. I turned it off in 3384081 to see if that sheds light on anything. So I wouldn't be surprised if it either disappeared or changed to a new error. Since every run I do with any number of cores or settings seems to work fine, LMK if anything changes on your end...

[funnell]

Well I've updated and run vastdiff three times without error, so maybe that did it!