gap-system / gap / 12025 / 9
4%
master: 4%

DEFAULT BRANCH: master
Ran
Files 640
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 59.771%. Remained the same
CFLAGS="--coverage -O2 -g" CXXFLAGS="--coverage -O2 -g" LDFLAGS="--coverage" CONFIGFLAGS="--enable-debug" COVERALLS_PARALLEL=true TEST_SUITES=testbugfix CONFIGFLAGS="--enable-memory-checking"

push

travis-ci

fingolfin 
buildsys: fix two regressions and add more tests

That first change modified $DEPFLAGS to ensure that .d files contain
targets for .lo files and not for the .d files -- the bug here was use
of the $@ variable which can have two values: the name of the .lo file
or the name of the .d file, depending on why the combined build rule for
the .lo/.d pair was triggered. This is easily fixed by using $OBJFILE
instead.

This bug meant that e.g. `make print-VAR` which should never lead to
anything being compiled could trigger compilation of stuff anyway.

Secondly, we want all *.lo files to depend on config.h; but the old way
to do that, which was to just let $(OBJS) depend on config.h, does not
work reliably anymore due to our combined .lo/.d rule: if the
compilation is triggered by the .d file being regenerated, we end up
calling the C compiler before config.h was regenerated. We could have
fixed this by adding a dependency from $(DEPFILES) to config.h, but it
seemed more natural to add config.h as a dependency to the rules which
actually invoke the compiler.

Finally we add a bunch of tests to `testbuildsys`, which ought to have
caught the two issues fixed above, plus some more.

168933 of 282635 relevant lines covered (59.77%)

2554393.5 hits per line

Source Files on job 12025.9 (CFLAGS="--coverage -O2 -g" CXXFLAGS="--coverage -O2 -g" LDFLAGS="--coverage" CONFIGFLAGS="--enable-debug" COVERALLS_PARALLEL=true TEST_SUITES=testbugfix CONFIGFLAGS="--enable-memory-checking")