J. Grant wrote: > On 03/01/06 12:51, Keith MARSHALL wrote: >> Eli Zaretskii wrote: >>> Do I understand correctly that you built the MinGW port of Make by >>> running "./configure; make" from the shell's prompt? If so, then >>> that's not how you should do it. Instead, run the build_w32.bat batch >>> file from cmd.exe, like this: >>> >>> build_w32 gcc >>> >>> (The file README.W32 explains this, I think.) >> >> Maybe, but under MSYS, MinGW builds are supposed to be accomplished >> by using a conventional >> >> ./configure && make > > That works fine for me, does it not work for you? I built make-3.81beta4
> the other day under MSYS and ran the testsuite. Yes, it works fine for me too; I cannot reproduce the OP's problem, with a conventional `./configure && make'; I see only one warning, repeated several times for glob.c, passing an incompatible pointer type for arg 0 to my_realloc, at line 197. Running the testsuite does throw up a number of reported failures. Examining the diff files suggests that quite a few of these are due to POSIX style path names appearing in the reference files, where the equivalent Win32 native path names appear in the test logs, so the failure reports may be spurious; several others are related to features, e.g. symlinks, which I would expect to fail in any case. OTOH, build.sh does not achieve a successful build under MSYS -- it quickly runs through compilation of ar.c and arscan.c, but then craps out with a "No such file or directory" error for w32err.h, while attempting to compile commands.c Regards, Keith. _______________________________________________ Make-w32 mailing list [email protected] http://lists.gnu.org/mailman/listinfo/make-w32
