[Bioc-devel] R CMD check without WARNING: g++ and STL issue for xcms and mzR
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 would it ? For now, just adding -O2 on Linux would fix this issue.
The code is compiled on the user machine, so twiddling with the build
system masks the problem from the developer while leaving the user
vulnerable.
Calling abort() is certainly a serious problem, abruptly ending the user
session.
Optimization could either eliminate a dead code path, or it could
compile the call to abort in a way that R no longer recognizes it -- R
is doing the equivalent of nm *.o |grep abort.
It would be appropriate to ignore the warning about abort() if it were
never reached; one would only know this by careful code analysis rather
than adjusting optimization flags.
Rsamtools re-defines abort (before including offending headers), and
avoids the warning (and crash), with the equivalent of
PKG_CFLAGS=-D__builtin_abort=_samtools_abort
where _samtools_abort is my own implementation to signal an R error
telling the user that an unrecoverable error has occurred and they
should stop what they are doing. Unfortunately this requires
-U_FORTIFY_SOURCE and doesn't actually address the reason for the abort,
and is hardly ideal.
I don't think the developer can set the optimization flag in a way that
overrides the R or user setting (the PKG_CXXFLAGS is inserted before the
R flags). And it doesn't address the underlying problem anyway.
Kasper's pragmatic approach (it's a very unlikely situation, and
difficult to fix, so live with the warning) seems like it would often be
a reasonable one.
FWIW this little bit of C++ seems to be enough to include abort
tmp.cpp:
-------
#include <list>
int foo(int argc, const char *argv[]) {
std::list<int> l1, l2;
std::list<int>::iterator it;
it = l1.begin();
l1.splice (it, l2); // mylist1: 1 10 20 30 2 3 4
return 0;
}
-------
Test with
rm -f tmp.o && R CMD SHLIB tmp.cpp && nm tmp.o | grep abort
with compiler settings in
~/.R/Makevars
-------------
CXXFLAGS = -g -O0
-------------
Martin
Yours, Steffen On Mi, 2017-04-12 at 10:07 -0400, Vincent Carey wrote:
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 will surely arise in other settings. Could BiocCheck prototype the approach? On Wed, Apr 12, 2017 at 9:12 AM, Kasper Daniel Hansen <kasperdanielha nsen at gmail.com> wrote:
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 to document that the included sources in the package are different from the official sources. I have had warnings about this in affxparser for a long, long time and no-one has bothered me over it. What might be nice is (somehow) setting a flag saying this should be ignored in the check process, but of course we don't want that flag to be set by the developer, because usually, these issues should be dealt with. So perhaps there is nothing to do and I always have to click on each build report to verify that there is only 1 warning from this issues and not others. Best, Kasper On Wed, Apr 12, 2017 at 7:22 AM, Neumann, Steffen <sneumann at ipb-hal le.de> wrote:
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 rid of the abort() symbol, as shown in https://github.com/sneumann/xcms/issues/150# issuecomment-293545521 => Is there a way to get -O2 back on the BioC build machines ? This should fix our WARNING. Bonus: will fix the same issue in
mzR :-)
Yours, Steffen On So, 2017-04-02 at 10:01 -0400, Martin Morgan wrote:
On 04/02/2017 06:52 AM, Neumann, Steffen wrote:
[...] 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
[...]
We are tracking this in: https://github.com/sneumann/xcms/issues/150
[...]
I don't think Bioconductor can help with this; maybe the Rcpp
or R-
devel mailing lists?
-- 3rd Leibniz Plant Biochemistry Symposium, 22.-23. of June 2017 http://www.ipb-halle.de/en/public-relations/3rd-leibniz-plant-bio
chemis
try-symposium/ -- IPB Halle AG Massenspektrometrie & Bioinformatik Dr. Steffen Neumann http://www.IPB-Halle.DE Weinberg 3 Tel. +49 (0) 345 5582 - 1470 06120 Halle +49 (0) 345 5582 - 0 sneumann(at)IPB-Halle.DE Fax. +49 (0) 345 5582 - 1409 -- 3rd Leibniz Plant Biochemistry Symposium, 22.-23. of June 2017 http://www.ipb-halle.de/en/public-relations/3rd-leibniz-plant-bio
chemis
try-symposium/ -- IPB Halle AG Massenspektrometrie & Bioinformatik Dr. Steffen Neumann http://www.IPB-Halle.DE Weinberg 3 Tel. +49 (0) 345 5582 - 1470 06120 Halle +49 (0) 345 5582 - 0 sneumann(at)IPB-Halle.DE Fax. +49 (0) 345 5582 - 1409
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
[[alternative HTML version deleted]]
_______________________________________________ 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}}