Re: [PATCH] bash-5.3-rc1 FTBFS when cross compiling with gcc-15
On 5/22/25 12:48 AM, Oğuz wrote: On Thursday, May 22, 2025, Bruce Dubbs wrote: with cstd-23 'bool' is a keyword This is something the configure script should check. Otherwise why have one? When you are cross-compiling, config.h is for the target system, not the build system. You need a conservative second file for the build system. -- ``The lyf so short, the craft so long to lerne.'' - Chaucer ``Ars longa, vita brevis'' - Hippocrates Chet Ramey, UTech, CWRUc...@case.eduhttp://tiswww.cwru.edu/~chet/ OpenPGP_signature.asc Description: OpenPGP digital signature
Re: [PATCH] bash-5.3-rc1 FTBFS when cross compiling with gcc-15
On 5/21/25 5:02 PM, Bruce Dubbs wrote: In linuxfromscratch we do a preliminary build of bash in order to run it later in a chroot environment without any dependencies from the host system: ./configure --prefix=/usr \ --build=x86_64-pc-linux-gnu \ --host=x86_64-lfs-linux-gnu \ --without-bash-malloc When we do this, the build runs: gcc -c -DHAVE_CONFIG_H -DSHELL -I. -I.. -I.. -I../include -I../lib -I. \ -Wno-parentheses -Wno-format-security -DDEBUG -g -DCROSS_COMPILING \ mkbuiltins.c Looking at ./builtins/mkbuiltins.c we have: #if !defined (CROSS_COMPILING) # include ... #include "../bashansi.h" Yes, cross-compiling means that config.h is for the target system, not the build system. The build system assumes a very conservative C90/XPG6 compilation environment. Bash-5.3 introduces a buildconf.h file for this situation; I'll see if I can fix this there and have mkbuiltins.c include it. -- ``The lyf so short, the craft so long to lerne.'' - Chaucer ``Ars longa, vita brevis'' - Hippocrates Chet Ramey, UTech, CWRUc...@case.eduhttp://tiswww.cwru.edu/~chet/ OpenPGP_signature.asc Description: OpenPGP digital signature
