Re: [Bioc-devel] R CMD check without WARNING: g++ and STL issue for xcms and mzR

2017-04-19 Thread Hervé Pagès
Thanks Steffen for bringing this to the R-devel list. I only see the fix in trunk for now. They'll need to port it to the 3.4 branch if we want to see it propagate to our BioC 3.5 builds. Just to clarify, by default R doesn't use the -Wall flag to compile packages on Linux. It's something we add

Re: [Bioc-devel] R CMD check without WARNING: g++ and STL issue for xcms and mzR

2017-04-19 Thread Neumann, Steffen
Hi, just a quick report, I followed this up to r-devel, and a fix adding -O2 back to default CXXFLAGS  has been committed (r72549, see [1]) to R by Martyn Plummer,  thus "fixing" the abort() related WARNING in mzR/xcms  once this gets into the R version used on malbec2. Yours, Steffen [1] 

Re: [Bioc-devel] R CMD check without WARNING: g++ and STL issue for xcms and mzR

2017-04-12 Thread Martin Morgan
On 04/12/2017 11:31 AM, Neumann, Steffen wrote: Hi, certainly these are good suggestions, but I would tend to simply add some whitelisting functionality if the cause is beyond the package's control. In this case I doubt it is a handler for SIGABRT, since that would not go away with -O2, or

Re: [Bioc-devel] R CMD check without WARNING: g++ and STL issue for xcms and mzR

2017-04-12 Thread Hervé Pagès
Hi Steffen, On 04/12/2017 04:22 AM, Neumann, Steffen wrote: Hi Martin and malbec2 admin(s): Some more digging revealed that malbec2

Re: [Bioc-devel] R CMD check without WARNING: g++ and STL issue for xcms and mzR

2017-04-12 Thread Neumann, Steffen
Hi, certainly these are good suggestions, but I would tend  to simply add some whitelisting functionality if the  cause is beyond the package's control.  In this case I doubt it is a handler for SIGABRT,  since that would not go away with -O2, or would it ? For now, just adding -O2 on Linux

Re: [Bioc-devel] R CMD check without WARNING: g++ and STL issue for xcms and mzR

2017-04-12 Thread Vincent Carey
Suppose you had a handler for SIGABRT in your code. Could CMD check check for that and, if found, refrain from warning? That is somewhat involved and goes beyond Bioc but it seems a principled way of dealing with operations in binary infrastructure whose behavior we cannot control. The problem

Re: [Bioc-devel] R CMD check without WARNING: g++ and STL issue for xcms and mzR

2017-04-12 Thread Kasper Daniel Hansen
I think "we" have to appreciate that the warning about abort/etc and others is really hard to deal with when you're including (large) external source as you do in mzR and for example affxparser / Rgraphviz. Generally fixing those in external code requires a lot of effort, and the further effort

Re: [Bioc-devel] R CMD check without WARNING: g++ and STL issue for xcms and mzR

2017-04-12 Thread Neumann, Steffen
Hi Martin and malbec2 admin(s): Some more digging revealed that malbec2  http://bioconductor.org/checkResults/devel/bioc-LATEST/malbec2-NodeInfo.html is using "CFLAGS=-g -O2 -Wall" but only "CXXFLAGS=-Wall" while e.g. tokay2 uses "CXXFLAGS=-O2 -Wall -mtune=core2" The -O2 optimisation is getting

[Bioc-devel] R CMD check without WARNING: g++ and STL issue for xcms and mzR

2017-04-02 Thread Neumann, Steffen
Hi, in preparation for the release, we are hunting down WARNINGS "Found ‘abort’, possibly from ‘abort’ (C)" in xcms/mzR. The abort() call is not coming from XCMS, but rather  from the C++ code in the STL, and we have no idea  how to get rid of it. We are tracking this in: