> Really looks like the infamous "include-fixed" problem. When some packages > are installed on the system, GCC may decide their include files are > "wrong", and fix them. Of course the first time GCC is built in chapter 6, > there are (almost) no packages on the system, and nothing is fixed. The > second time you build gcc (using the first one), it may happen that > packages "to be fixed" (according to GCC) are now present, so that they are > fixed... Whether this has some relationship with your problem, I am not > sure. But it is at least one way to explain a different behavior at the > second GCC compilation.
To mee, it look like that (include_fixed) and upon reading the message thread here and some googling, I do think both are related. I haven't finished reading that post there: https://developers.redhat.com/blog/2016/02/29/why-cstdlib-is-more-complicated-than-you-might-think/ but in my case, BLFS packages installation are fixed using these environment variables: LC_ALL=POSIX LANG=en_CA.UTF8 JAVA_HOME=/opt/jdk HOME=/home/lfs XORG_PREFIX=/usr XORG_CONFIG="--prefix=/usr --sysconfdir=/etc --localstatedir=/var --disable- static" SHELL=/bin/bash GIO_USE_TLS=gnutls-pkcs11 PATH=/bin:/usr/bin:/opt/jdk/bin:/opt/maven/bin:/opt/texlive/2018/bin/x86_64- linux some of the PATH entries are provisionals (to be installed but not yet installed) but I set them anyway to prevent breakage of compilation cycle later on when the package actually get installed. No other variable get exported on this account and I do not have any issues with include_fixed directives or any other C++ related errors. Alain -- http://lists.linuxfromscratch.org/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
