On Wed, Apr 17, 2019 at 11:28:56AM +0500, ???? ??????? wrote: > ??, 17 ???. 2019 ?. ? 07:29, Willy Tarreau <[email protected]>: > > > Hi Ilya, > > > > On Tue, Apr 16, 2019 at 11:17:49PM +0500, ???? ??????? wrote: > > > +env: > > > + global: > > > + - USE_THREAD=1 > > > + - USE_OPENSSL=1 > > > + - USE_PCRE=1 > > > + - USE_ZLIB=1 > > > + - USE_GETADDRINFO=1 > > > > It's unclear to me how these ones are supposed to be used by the Makefile > > considering that they are not passed on the command line; thus the target's > > default values will still apply. > > > > those are used by Makefile. however, I agree with you that it is not common > way of passing variables.
I know these are *expected* to be used by the make file, what I mean is that if they are only set as environment variables, they have less precedence than defaults set in the makefile. > > > +matrix: > > > + include: > > > +# - os: linux > > > +# compiler: gcc > > > +# env: TARGET=linux2628 > > > + - os: linux > > > + compiler: clang > > > + env: TARGET=linux2628 > > > +# - os: osx > > > +# compiler: clang > > > +# env: TARGET=osx SSL_LIB=/usr/local/opt/openssl/lib > > SSL_INC=/usr/local/opt/openssl/include TMPDIR=/var/tmp > > > + > > > +install: > > > + - git clone https://github.com/VTest/VTest.git > > > + - make -C VTest > > > > I've seen on your previous test report page that vtest fails to build > > with the gcc version there due to some zealous checks on the return > > value of "nice()" which expect to be verified in the include files. > > I'd suggest to simply force the build options so that it's never an > > issue and that we don't fail on other uninteresting errors, since > > vtest is a utility and not the primary component we're testing : > > > > - make -C VTest FLAGS='-O2 -s -Wall -Wno-unused-result' > > > > well, I hope it will be resolved in VTest itself. I opened bug there > > https://github.com/vtest/VTest/issues/12 > > otherwise we will aplly some hack. let us wait for a while. OK, fair enough. > > > +script: > > > + - make CC=$CC V=1 TARGET=$TARGET > > > > So here you should pass your global options, since by default they will > > not be used if explicitly disabled, or not be disabled if explicitly > > enabled : > > > > - make CC=$CC V=1 TARGET=$TARGET USE_THREAD=$USE_THREAD > > USE_OPENSSL=$USE_OPENSSL USE_PCRE=$USE_PCRE USE_ZLIB=$USE_ZLIB > > USE_GETADDRINFO=$USE_GETADDRINFO > > > > > the main issue with build matrix is regression. > for example build fails without threads enabled. > > we will adopt matrix I'm not sure what you mean or how it is related to the comment above. The only think I meant was to *force* the variables on make so as not to rely on unknown defaults. Regards, Willy
