Re: svn commit: r297435 - head: still problems for stage 3 when gcc 4.2.1 is avoided (powerpc64 self-hosted build)

From: Mark Millard <markmi_at_dsl-only.net>
Date: Sat, 2 Apr 2016 23:06:48 -0700
On 2016-Apr-2, at 3:59 PM, Mark Millard <markmi at dsl-only.net> wrote:

> [My testing for the likes of the below does not yet extend outside powerpc64 contexts.]
> 
> For the likes of self-hosted powerpc64-xtoolchain-gcc/powerpc64-gcc use with, say, gcc49 materials as the so-called "host" compiler tools I have not yet found a way around using the workaround:
> 
>> # ls -l /usr/lib/libstdc++.*
>> lrwxr-xr-x  1 root  wheel  17 Feb 23 00:09 /usr/lib/libstdc++.a -> /usr/lib/libc++.a
>> lrwxr-xr-x  1 root  wheel  18 Feb 23 00:09 /usr/lib/libstdc++.so -> /usr/lib/libc++.so
> 
> 
> 
> But I appear to now have a src.conf (or a family of them) that avoids needing workarounds in /usr/local/include and /usr/local/lib for filename conflicts. This is based on CC/CXX ("HOST") and XCC/XCXX ("CROSS") in src.conf being the likes of:
> 
> "HOST" (CC/CXX):
>> CC=env C_INCLUDE_PATH=/usr/include /usr/local/bin/gcc49 -L/usr/lib
>> CXX=env C_INCLUDE_PATH=/usr/include CPLUS_INCLUDE_PATH=/usr/include/c++/v1 /usr/local/bin/g++49 -std=c++11 -nostdinc++ -L/usr/lib
> 
> and. . .
> 
> "CROSS" (XCC/XCXX):
>> TO_TYPE=powerpc64
>> TOOLS_TO_TYPE=${TO_TYPE}
>> . . .
>> VERSION_CONTEXT=11.0
>> . . .
>> XCC=/usr/local/bin/${TOOLS_TO_TYPE}-portbld-freebsd${VERSION_CONTEXT}-gcc
>> XCXX=/usr/local/bin/${TOOLS_TO_TYPE}-portbld-freebsd${VERSION_CONTEXT}-g++
> 
> In other words: CROSS use is already handled for /usr/local/. . . just given the compiler paths but some special src.conf adjustments beyond compiler paths were needed for HOST.
> 
> So far it appears that gcc49 materials can be used in "CROSS" and/or powerpc64-gcc materials can be used in "HOST" via just appropriate compiler-path editing. (I have jumped to testing -r297514 but am still doing various build tests. So this aspect is subject to updates.) It appears to be possible to use just one compiler/tool family --but in the 2 different ways-- instead of using a mix of 2 compiler/tool families.
> 
> Historically I've not gotten gcc variants to make a working lib32 for powerpc64 and I've yet to retest this: So no claims about the lib32 status are implied here. (The problem was code in crtbeginS that arbitrarily used R30 in a way that the context was not set up for and so crtbeginS code was dereferencing arbitrary addresses.)

I probably knew this someplace in the back of my head but gcc49 does not handle -fformat-extensions (quoting the script log):

> --- accf_data.o ---
> env /usr/local/bin/gcc49 -isystem /usr/obj/xtoolchain/powerpc.powerpc64/usr/src/tmp/usr/include -L/usr/obj/xtoolchain/powerpc.powerpc64/usr/src/tmp/usr/lib --sysroot=/usr/obj/xtoolchain/powerpc.powerp
> c64/usr/src/tmp -B/usr/local/powerpc64-portbld-freebsd11.0/bin/ -O2 -pipe  -fno-strict-aliasing -Werror -D_KERNEL -DKLD_MODULE -nostdinc   -DHAVE_KERNEL_OPTION_HEADERS -include /usr/obj/xtoolchain/powerpc.powerpc64/usr/src/sys/GENERIC64vtsc-NODEBUG/opt_global.h -I. -I/usr/src/sys -fno-common -g -mlongcall -fno-omit-frame-pointer -I/usr/obj/xtoolchain/powerpc.powerpc64/usr/src/sys/GENERIC64vtsc-NODEBUG  -MD  -MF.depend.accf_data.o -MTaccf_data.o -mno-altivec -ffreestanding -fwrapv -fstack-protector -gdwarf-2 -Wall -Wredundant-decls -Wnested-externs -Wstrict-prototypes  -Wmissing-prototypes -Wpointer-arith -Winline -Wcast-qual  -Wundef -Wno-pointer-sign -fformat-extensions  -Wmissing-include-dirs -fdiagnostics-show-option  -Wno-unknown-pragmas  -Wno-error=inline -Wno-error=enum-compare -Wno-error=unused-but-set-variable  -Wno-error=aggressive-loop-optimizations -Wno-error=maybe-uninitialized  -Wno-error=array-bounds -Wno-error=address  -Wno-error=cast-qual -Wno-error=sequence-point -Wno-error=attributes  -Wno-error=strict-overflow -Wno-error=overflow  -v -finline-limit=15000 -fms-extensions --param inline-unit-growth=100 --param large-function-growth=1000 -msoft-float -mcall-aixdesc -std=iso9899:1999 -c /usr/src/sys/modules/accf_data/../../netinet/accf_data.c -o accf_data.o
> Using built-in specs.
> COLLECT_GCC=/usr/local/bin/gcc49
> gcc49: error: unrecognized command line option '-fformat-extensions'
> Target: powerpc64-portbld-freebsd11.0
> Configured with: ./../gcc-4.9-20160210/configure --disable-multilib --disable-bootstrap --disable-nls --enable-gnu-indirect-function --libdir=/usr/local/lib/gcc49 --libexecdir=/usr/local/libexec/gcc49 --program-suffix=49 --with-as=/usr/local/bin/as --with-gmp=/usr/local --with-gxx-include-dir=/usr/local/lib/gcc49/include/c++/ --with-ld=/usr/local/bin/ld --with-pkgversion='FreeBSD Ports Collection' --with-system-zlib --disable-libgcj --enable-languages=c,c++,objc,fortran --prefix=/usr/local --localstatedir=/var --mandir=/usr/local/man --infodir=/usr/local/info/gcc49 --build=powerpc64-portbld-freebsd11.0
> Thread model: posix
> gcc version 4.9.4 20160210 (prerelease) (FreeBSD Ports Collection) 
> *** [accf_data.o] Error code 1

So my

> it appears that gcc49 materials can be used in "CROSS"

is just false for gcc49, gcc5, and the like.

I had hoped such would work with TARGET_ARCH=powerpc because there is no powerpc-gcc port predefined and clang 3.8.0 is still insufficient for this context.

===
Mark Millard
markmi at dsl-only.net

On 2016-Apr-1, at 4:35 PM, Mark Millard <markmi at dsl-only.net> wrote:
> 
> [Just a top-post showing what powerpc64-xtoolchain-gcc/powerpc64-gcc has for the default include search places:]
> 
> powerpc64-xtoolchain-gcc/powerpc64-gcc also looks in /usr/local/include before /usr/include : see below.
> 
>> # portmaster --list-origins
>> . . .
>> devel/powerpc64-xtoolchain-gcc
>> . . .
>> 
>> # /usr/local/bin/powerpc64-portbld-freebsd11.0-gcc --version
>> powerpc64-portbld-freebsd11.0-gcc (FreeBSD Ports Collection for powerpc64) 5.3.0
>> Copyright (C) 2015 Free Software Foundation, Inc.
>> This is free software; see the source for copying conditions.  There is NO
>> warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
>> 
>> # echo '' |/usr/local/bin/powerpc64-portbld-freebsd11.0-gcc -v -x c++ - -o /dev/null
>> . . .
>> ignoring nonexistent directory "/usr/local/lib/gcc/powerpc64-portbld-freebsd11.0/5.3.0/../../../../include/c++/5.3.0"
>> ignoring nonexistent directory "/usr/local/lib/gcc/powerpc64-portbld-freebsd11.0/5.3.0/../../../../include/c++/5.3.0/powerpc64-portbld-freebsd11.0"
>> ignoring nonexistent directory "/usr/local/lib/gcc/powerpc64-portbld-freebsd11.0/5.3.0/../../../../include/c++/5.3.0/backward"
>> ignoring nonexistent directory "/usr/local/lib/gcc/powerpc64-portbld-freebsd11.0/5.3.0/../../../../powerpc64-portbld-freebsd11.0/include"
>> #include "..." search starts here:
>> #include <...> search starts here:
>> /usr/local/lib/gcc/powerpc64-portbld-freebsd11.0/5.3.0/include
>> /usr/local/include
>> /usr/local/lib/gcc/powerpc64-portbld-freebsd11.0/5.3.0/include-fixed
>> /usr/include
>> End of search list.
>> . . .
> 
> 
> ===
> Mark Millard
> markmi at dsl-only.net

> On 2016-Apr-1, at 7:25 AM, Warner Losh <imp at bsdimp.com> wrote:
> 
> 
> 
> On Fri, Apr 1, 2016 at 2:25 AM, Dimitry Andric <dim_at_freebsd.org> wrote:
> On 01 Apr 2016, at 00:44, Warner Losh <imp_at_bsdimp.com> wrote:
>> 
>>> On Mar 31, 2016, at 4:34 PM, Bryan Drewery <bdrewery_at_freebsd.org> wrote:
>>> I didn't realize the ports compiler was defaulting /usr/local/include
>>> into the search path now.  It does not have /usr/local/lib in the
>>> default library path as far as I can tell.  It's also broken for its
>>> -rpath (noted in its pkg-message).  So having a default
>>> /usr/local/include path seems odd.
>> 
>> It has for a while now. It’s one of the maddening inconsistencies that abound in this
>> area. I took a poll a while ago and there seemed to be widespread support for adding
>> it to the base compiler.
> 
> This was the main reason /usr/local/include was *not* included in the
> base compiler, otherwise it would unpredictably pick up headers in
> /usr/local/include during builds.  You can never know which conflicting
> headers a certain user has installed in /usr/local/include... :)
> 
> That's why it shouldn't be *FIRST*, not why it shouldn't be there
> at all.
> 
>>> Adding -isystem /usr/include to fix this is probably possible but
>>> there's a risk someone will remove it as redundant.  In this case I wish
>>> /usr/include was first but I'm not sure what impact that would have on
>>> consumers expecting /usr/local/include (and /usr/local/lib) overrides to
>>> work, though they would need to pass a -L /usr/local/lib anyhow and
>>> would likely be passing -I /usr/local/lib too.
>> 
>> /usr/include should be first. But it isn’t. That’s another inconsistency that was found
>> when we looked at /usr/local stuff. Someone recently added /usr/local/bin to the path,
>> if I recall correctly.
> 
> Isn't that a bit of a bikeshed?  I guess some people would just as well
> prefer /usr/local/include to be first, just like some people prefer
> /usr/local/bin before /usr/bin in their PATH.
> 
> Sigh. No. /usr/local is moving into many different things in the base and ports. We should
> do it in a consistent way. What we have now is not consistent and somewhat brittle.
> 
> In any case, if such paths are added to external compilers, we better
> make sure almost everything in buildworld uses -nostdinc, and specifying
> exactly the include directories we need, and no others.  Cumbersome, but
> maybe for a good cause.
> 
> That's the non-brittle solution here. An over-reliance on defaults makes it
> difficult to ensure other compilers will work, especially ones we don't
> tightly control the defaults for.
> 
> Warner
Received on Sun Apr 03 2016 - 05:02:30 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:41:03 UTC