[ https://issues.apache.org/jira/browse/TS-1091?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13191429#comment-13191429 ]
Marc Abramowitz commented on TS-1091: ------------------------------------- Or a more extreme change would be to use an existing autoconf macro for this: http://www.gnu.org/software/autoconf-archive/ax_func_which_gethostbyname_r.html This one takes a different approach - it does a sanity check where it tries to compile with a one arg gethostbyname_r which doesn't exist so if that passes, then it knows something is wrong. > `./configure CFLAGS=-w` causes configure script to wrongly guess style of > `gethostbyname_r` on OS X (and probably other BSDs) > ----------------------------------------------------------------------------------------------------------------------------- > > Key: TS-1091 > URL: https://issues.apache.org/jira/browse/TS-1091 > Project: Traffic Server > Issue Type: Bug > Components: Build > Affects Versions: 3.0.2 > Environment: OS X 10.6.8 > Reporter: Marc Abramowitz > Priority: Minor > Labels: autoconf, build, configure > Original Estimate: 10m > Remaining Estimate: 10m > > {code} > marca@SCML-MarcA:~/src/trafficserver-3.0.2$ system_profiler -detailLevel mini > | grep 'System Version' > System Version: Mac OS X 10.6.8 (10K549) > marca@SCML-MarcA:~/src/trafficserver-3.0.2$ ./configure CFLAGS=-w > ... > checking style of gethostbyname_r routine... glibc2 > checking 3rd argument to the gethostbyname_r routines... hostent_data > configure: Build using CC=gcc > configure: Build using CXX=g++ > configure: Build using CPP=gcc -E > configure: Build using CFLAGS=-w -g -pipe -Wall -Werror -O3 > -feliminate-unused-debug-symbols -fno-strict-aliasing > ... > marca@SCML-MarcA:~/src/trafficserver-3.0.2$ grep gethostbyname_r Makefile > gethostbyname_r_glibc2 = 1 > gethostbyname_r_hostent_data = 1 > marca@SCML-MarcA:~/src/trafficserver-3.0.2$ make > ... > ink_inet.cc: In function 'hostent* ink_gethostbyaddr_r(char*, int, int, > ink_gethostbyaddr_r_data*)': > ink_inet.cc:73: error: cannot convert 'hostent**' to 'int*' for argument '7' > to 'hostent* gethostbyaddr_r(const char*, size_t, int, hostent*, char*, int, > int*)' > make[3]: *** [ink_inet.lo] Error 1 > make[2]: *** [all] Error 2 > make[1]: *** [all-recursive] Error 1 > make: *** [all-recursive] Error 1 > {code} > Arguably, people should never run configure with CFLAGS=-w and this might > seem stupid and something that should never happen, but it turns out that > Homebrew (http://mxcl.github.com/homebrew/) does this by default > (https://github.com/mxcl/homebrew/issues/9728) -- I discovered this while > writing a Homebrew formula for Traffic Server > (https://github.com/mxcl/homebrew/pull/9513). After discovering that this was > causing problems, I modified my formula to tell Homebrew not to do this and > all is well, but I thought it would be interesting to make Traffic Server > resistant to this as well. > I first tried to enable warnings by trying to find a gcc #pragma that could > go in the conftest.c code > (http://gcc.gnu.org/onlinedocs/gcc/Diagnostic-Pragmas.html looked promising > but I could not find one that worked), but I could not find any #pragma that > seemed to do this. > I ended up making a small change to `build/common.m4` that strips -w out of > CFLAGS using `sed`. > {code} > --- build/common.m4.2012-01-22-092051 2011-05-25 13:19:54.000000000 -0700 > +++ build/common.m4 2012-01-22 22:48:14.000000000 -0800 > @@ -177,6 +177,7 @@ > if test "$ac_cv_prog_gcc" = "yes"; then > CFLAGS="$CFLAGS -Werror" > fi > + CFLAGS=$(echo $CFLAGS | sed -e 's/^-w$//' -e 's/^-w //' -e 's/ -w$//' -e > 's/ -w / /') > AC_COMPILE_IFELSE([AC_LANG_SOURCE([ > [#include "confdefs.h" > ] > {code} -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira