Unset macros to prevent 'redefined' errors during build

[x4121]

In preparation to a (hopefully soon :) ) new release of newsbeuter, I tried to build Debian/Ubuntu binaries. Currently the build fails with

<command-line>:0:0: error: "LOCALEDIR" redefined [-Werror]
<command-line>:0:0: note: this is the location of the previous definition
cc1plus: all warnings being treated as errors
Makefile:112: recipe for target 'test/cache.o' failed
make[1]: *** [test/cache.o] Error 1

Unsetting the build macros before (re)setting them prevents this.

Tested on my private Launchpad PPA: https://launchpad.net/~x4121/+archive/ubuntu/ppa/+build/12464800 (which should behave similar/identical to the Debian build server that the release would be built on)

[coveralls]

Coverage remained the same at 36.726% when pulling 3370194 on x4121:pr/debuild_prepare into 5517dba on akrennmair:master.

[coveralls]

Coverage remained the same at 36.726% when pulling 3370194 on x4121:pr/debuild_prepare into 5517dba on akrennmair:master.

[Minoru]

This might be related to #431.

Unsetting macros is more of a workaround than a fix. Anyway, I'd like to know what causes this before deciding how to deal with it. What parameters are passed to make? What's the environment like (env)?

Debian package is maintained by @tsipinakis. Maybe he's got some information on this?

[x4121]

Ah you are right.. I thought the problem was the macros being set during make and again during make test afterwards, because it only complained there..

All build steps look like (manually inserted line breaks for readability)

g++ -g -O2 -fPIE -fstack-protector-strong -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2
-std=c++11 -ggdb -Iinclude -Istfl -Ifilter -I. -Irss -Werror -Wall -Wextra -Wunreachable-code
-DLOCALEDIR=\"/usr/share/locale\" -I/usr/include/libxml2  -I/usr/include/json-c  -I/usr/include/ncursesw 
-g -O2 -fPIE -fstack-protector-strong -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2
-std=c++11 -ggdb -Iinclude -Istfl -Ifilter -I. -Irss -Werror -Wall -Wextra -Wunreachable-code
-DLOCALEDIR=\"/usr/share/locale\" -o src/configcontainer.o -c src/configcontainer.cpp

So yes, the macros get defined twice every step, but fails at

g++ -g -O2 -fPIE -fstack-protector-strong -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2
-std=c++11 -ggdb -Iinclude -Istfl -Ifilter -I. -Irss -Werror -Wall -Wextra -Wunreachable-code
-DLOCALEDIR=\"/usr/local/share/locale\" -I/usr/include/libxml2  -I/usr/include/json-c 
-I/usr/include/ncursesw  -g -O2 -fPIE -fstack-protector-strong -Wformat -Werror=format-security
-D_FORTIFY_SOURCE=2 -std=c++11 -ggdb -Iinclude -Istfl -Ifilter -I. -Irss -Werror -Wall -Wextra
-Wunreachable-code -DLOCALEDIR=\"/usr/share/locale\" -o test/cache.o -c test/cache.cpp
<command-line>:0:0: error: "LOCALEDIR" redefined [-Werror]
<command-line>:0:0: note: this is the location of the previous definition

[tsipinakis]

I was in the middle of dumping a wall of text with all the variables that are set during the build process but I figured it out in the meantime. It's the same issue as #431.

If config.mk doesn't exist, config.sh is called by make which adds CXXFLAGS+=$CXXFLAGS to config.mk which is intended to append user-defined overrides to the flags, but make by default does not export its own variables unless they have been already set, in which case, it updates the values with whatever modifications happened within the makefile.

So if CXXFLAGS is not set, it's not exported so config.mk does not contain an extra copy of the flags and the bug does not happen. But if it is, it is updated with all the extra flags added inside the makefile, which results in 2 copies of CXXFLAGS, one from the makefile and one from the config.mk inclusion.

[Minoru]

@x4121, notice how the first command has two LOCALEDIR definitions with equal values, but the second has two definitions with different values. I wonder what's going on there. It looks like make test is called with a slightly different set of parameters than make. Can you look further into the script to see what might cause this?

@tsipinakis, thanks a lot for figuring this out! Let's move to #431 though.

[x4121]

Totally missed the "local" in the second definition...
Will try to have a deeper look into it on the weekend :)

[Minoru]

Hi,

I'm going to close this because I'm not going to merge this changes, but if you find anything while looking for that second "LOCALEDIR" definition, please open an issue.