| Message ID | 1470672688-6754-1-git-send-email-peter.maydell@linaro.org |
|---|---|
| State | New |
| Headers | show |
On 8 August 2016 at 17:11, Peter Maydell <peter.maydell@linaro.org> wrote: > The various host OSes are irritatingly variable about the name > of the linker emulation we need to pass to ld's -m option to > build the i386 option ROMs. Instead of doing this via a > CONFIG ifdef, check in configure whether any of the emulation > names we know about will work and pass the right answer through > to the makefile. If we can't find one, we fall back to not trying > to build the option ROMs, in the same way we would for a non-x86 > host platform. > > This is in particular necessary to unbreak the build on OpenBSD, > since it wants a different answer to FreeBSD and we don't have > an existing CONFIG_ variable that distinguishes the two. > > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > --- > This works for Linux and for the Windows builds; I don't have any > BSD systems to test it on. Brad and Sean, can I ask you to test this > on OpenBSD and FreeBSD, please? > > (I think this is going to miss -rc2. Sorry about that; we'll > get OpenBSD builds fixed for -rc3.) PS: if you would like your BSD flavour promoted to "we don't merge changes that break the build on it" I would need to have ssh access to a machine I can run builds and tests on. (Unfortunately the only BSD in the gcc compile farm is an ancient NetBSD.) thanks -- PMM
On 08/08/16 10:11, Peter Maydell wrote: > The various host OSes are irritatingly variable about the name > of the linker emulation we need to pass to ld's -m option to > build the i386 option ROMs. Instead of doing this via a > CONFIG ifdef, check in configure whether any of the emulation > names we know about will work and pass the right answer through > to the makefile. If we can't find one, we fall back to not trying > to build the option ROMs, in the same way we would for a non-x86 > host platform. > > This is in particular necessary to unbreak the build on OpenBSD, > since it wants a different answer to FreeBSD and we don't have > an existing CONFIG_ variable that distinguishes the two. > Reviewed-by: Sean Bruno <sbruno@freebsd.org> For impact to FreeBSD only. > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > --- > This works for Linux and for the Windows builds; I don't have any > BSD systems to test it on. Brad and Sean, can I ask you to test this > on OpenBSD and FreeBSD, please? > > (I think this is going to miss -rc2. Sorry about that; we'll > get OpenBSD builds fixed for -rc3.) > > configure | 12 +++++++++++- > pc-bios/optionrom/Makefile | 12 +----------- > 2 files changed, 12 insertions(+), 12 deletions(-) > > diff --git a/configure b/configure > index f57fcc6..7c744ad 100755 > --- a/configure > +++ b/configure > @@ -4699,7 +4699,16 @@ roms= > if test \( "$cpu" = "i386" -o "$cpu" = "x86_64" \) -a \ > "$targetos" != "Darwin" -a "$targetos" != "SunOS" -a \ > "$softmmu" = yes ; then > - roms="optionrom" > + # Different host OS linkers have different ideas about the name of the ELF > + # emulation. Linux and OpenBSD use 'elf_i386'; FreeBSD uses the _fbsd > + # variant; and Windows uses i386pe. > + for emu in elf_i386 elf_i386_fbsd i386pe; do > + if "$ld" -verbose 2>&1 | grep -q "^[[:space:]]*$emu[[:space:]]*$"; then > + ld_i386_emulation="$emu" > + roms="optionrom" > + break > + fi > + done > fi > if test "$cpu" = "ppc64" -a "$targetos" != "Darwin" ; then > roms="$roms spapr-rtas" > @@ -5539,6 +5548,7 @@ fi > echo "LDFLAGS=$LDFLAGS" >> $config_host_mak > echo "LDFLAGS_NOPIE=$LDFLAGS_NOPIE" >> $config_host_mak > echo "LD_REL_FLAGS=$LD_REL_FLAGS" >> $config_host_mak > +echo "LD_I386_EMULATION=$ld_i386_emulation" >> $config_host_mak > echo "LIBS+=$LIBS" >> $config_host_mak > echo "LIBS_TOOLS+=$libs_tools" >> $config_host_mak > echo "PTHREAD_LIB=$PTHREAD_LIB" >> $config_host_mak > diff --git a/pc-bios/optionrom/Makefile b/pc-bios/optionrom/Makefile > index 24e175e..5bbe233 100644 > --- a/pc-bios/optionrom/Makefile > +++ b/pc-bios/optionrom/Makefile > @@ -41,18 +41,8 @@ build-all: multiboot.bin linuxboot.bin linuxboot_dma.bin kvmvapic.bin > %.o: %.S > $(call quiet-command,$(CPP) $(QEMU_INCLUDES) $(QEMU_DGFLAGS) -c -o - $< | $(AS) $(ASFLAGS) -o $@," AS $(TARGET_DIR)$@") > > -ifdef CONFIG_WIN32 > -LD_EMULATION = i386pe > -else > -ifdef CONFIG_BSD > -LD_EMULATION = elf_i386_fbsd > -else > -LD_EMULATION = elf_i386 > -endif > -endif > - > %.img: %.o > - $(call quiet-command,$(LD) $(LDFLAGS_NOPIE) -m $(LD_EMULATION) -Ttext 0 -e _start -s -o $@ $<," Building $(TARGET_DIR)$@") > + $(call quiet-command,$(LD) $(LDFLAGS_NOPIE) -m $(LD_I386_EMULATION) -Ttext 0 -e _start -s -o $@ $<," Building $(TARGET_DIR)$@") > > %.raw: %.img > $(call quiet-command,$(OBJCOPY) -O binary -j .text $< $@," Building $(TARGET_DIR)$@") >
On 08/08/2016 19:12, Peter Maydell wrote: > PS: if you would like your BSD flavour promoted to "we don't > merge changes that break the build on it" I would need to > have ssh access to a machine I can run builds and tests on. > (Unfortunately the only BSD in the gcc compile farm is an > ancient NetBSD.) For the purpose of "not breaking the build", having a docker image for a cross compiler would also be enough. Paolo
On 9 August 2016 at 14:12, Paolo Bonzini <pbonzini@redhat.com> wrote: > On 08/08/2016 19:12, Peter Maydell wrote: >> PS: if you would like your BSD flavour promoted to "we don't >> merge changes that break the build on it" I would need to >> have ssh access to a machine I can run builds and tests on. >> (Unfortunately the only BSD in the gcc compile farm is an >> ancient NetBSD.) > > For the purpose of "not breaking the build", having a docker image for a > cross compiler would also be enough. My requirement here is largely that I personally don't have to expend any effort on admining and maintaining BSD related build setups. I also think that setups that let you run "make check" are much more valuable than cross-compile setups that don't. thanks -- PMM
On 08/08/16 12:11, Peter Maydell wrote: > The various host OSes are irritatingly variable about the name > of the linker emulation we need to pass to ld's -m option to > build the i386 option ROMs. Instead of doing this via a > CONFIG ifdef, check in configure whether any of the emulation > names we know about will work and pass the right answer through > to the makefile. If we can't find one, we fall back to not trying > to build the option ROMs, in the same way we would for a non-x86 > host platform. > > This is in particular necessary to unbreak the build on OpenBSD, > since it wants a different answer to FreeBSD and we don't have > an existing CONFIG_ variable that distinguishes the two. To add to this.. amd64 uses "elf_i386". Looking at i386 it uses elf_i386_obsd and that is the only emulation supported. So change the list of emulations to try to elf_i386 elf_i386_fbsd elf_i386_obsd i386pe Someone needs to take a look at what NetBSD / DragonFly are using. > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > --- > This works for Linux and for the Windows builds; I don't have any > BSD systems to test it on. Brad and Sean, can I ask you to test this > on OpenBSD and FreeBSD, please? > > (I think this is going to miss -rc2. Sorry about that; we'll > get OpenBSD builds fixed for -rc3.) > > configure | 12 +++++++++++- > pc-bios/optionrom/Makefile | 12 +----------- > 2 files changed, 12 insertions(+), 12 deletions(-) > > diff --git a/configure b/configure > index f57fcc6..7c744ad 100755 > --- a/configure > +++ b/configure > @@ -4699,7 +4699,16 @@ roms= > if test \( "$cpu" = "i386" -o "$cpu" = "x86_64" \) -a \ > "$targetos" != "Darwin" -a "$targetos" != "SunOS" -a \ > "$softmmu" = yes ; then > - roms="optionrom" > + # Different host OS linkers have different ideas about the name of the ELF > + # emulation. Linux and OpenBSD use 'elf_i386'; FreeBSD uses the _fbsd > + # variant; and Windows uses i386pe. > + for emu in elf_i386 elf_i386_fbsd i386pe; do > + if "$ld" -verbose 2>&1 | grep -q "^[[:space:]]*$emu[[:space:]]*$"; then > + ld_i386_emulation="$emu" > + roms="optionrom" > + break > + fi > + done > fi > if test "$cpu" = "ppc64" -a "$targetos" != "Darwin" ; then > roms="$roms spapr-rtas" > @@ -5539,6 +5548,7 @@ fi > echo "LDFLAGS=$LDFLAGS" >> $config_host_mak > echo "LDFLAGS_NOPIE=$LDFLAGS_NOPIE" >> $config_host_mak > echo "LD_REL_FLAGS=$LD_REL_FLAGS" >> $config_host_mak > +echo "LD_I386_EMULATION=$ld_i386_emulation" >> $config_host_mak > echo "LIBS+=$LIBS" >> $config_host_mak > echo "LIBS_TOOLS+=$libs_tools" >> $config_host_mak > echo "PTHREAD_LIB=$PTHREAD_LIB" >> $config_host_mak > diff --git a/pc-bios/optionrom/Makefile b/pc-bios/optionrom/Makefile > index 24e175e..5bbe233 100644 > --- a/pc-bios/optionrom/Makefile > +++ b/pc-bios/optionrom/Makefile > @@ -41,18 +41,8 @@ build-all: multiboot.bin linuxboot.bin linuxboot_dma.bin kvmvapic.bin > %.o: %.S > $(call quiet-command,$(CPP) $(QEMU_INCLUDES) $(QEMU_DGFLAGS) -c -o - $< | $(AS) $(ASFLAGS) -o $@," AS $(TARGET_DIR)$@") > > -ifdef CONFIG_WIN32 > -LD_EMULATION = i386pe > -else > -ifdef CONFIG_BSD > -LD_EMULATION = elf_i386_fbsd > -else > -LD_EMULATION = elf_i386 > -endif > -endif > - > %.img: %.o > - $(call quiet-command,$(LD) $(LDFLAGS_NOPIE) -m $(LD_EMULATION) -Ttext 0 -e _start -s -o $@ $<," Building $(TARGET_DIR)$@") > + $(call quiet-command,$(LD) $(LDFLAGS_NOPIE) -m $(LD_I386_EMULATION) -Ttext 0 -e _start -s -o $@ $<," Building $(TARGET_DIR)$@") > > %.raw: %.img > $(call quiet-command,$(OBJCOPY) -O binary -j .text $< $@," Building $(TARGET_DIR)$@") >
On 08/08/16 12:11, Peter Maydell wrote: > The various host OSes are irritatingly variable about the name > of the linker emulation we need to pass to ld's -m option to > build the i386 option ROMs. Instead of doing this via a > CONFIG ifdef, check in configure whether any of the emulation > names we know about will work and pass the right answer through > to the makefile. If we can't find one, we fall back to not trying > to build the option ROMs, in the same way we would for a non-x86 > host platform. > > This is in particular necessary to unbreak the build on OpenBSD, > since it wants a different answer to FreeBSD and we don't have > an existing CONFIG_ variable that distinguishes the two. > > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > --- > This works for Linux and for the Windows builds; I don't have any > BSD systems to test it on. Brad and Sean, can I ask you to test this > on OpenBSD and FreeBSD, please? > > (I think this is going to miss -rc2. Sorry about that; we'll > get OpenBSD builds fixed for -rc3.) Tested on OpenBSD/amd64 and it builds. Need to get my i386 VM going and test on i386.
On 11 August 2016 at 00:42, Brad Smith <brad@comstyle.com> wrote: > On 08/08/16 12:11, Peter Maydell wrote: >> >> The various host OSes are irritatingly variable about the name >> of the linker emulation we need to pass to ld's -m option to >> build the i386 option ROMs. Instead of doing this via a >> CONFIG ifdef, check in configure whether any of the emulation >> names we know about will work and pass the right answer through >> to the makefile. If we can't find one, we fall back to not trying >> to build the option ROMs, in the same way we would for a non-x86 >> host platform. >> >> This is in particular necessary to unbreak the build on OpenBSD, >> since it wants a different answer to FreeBSD and we don't have >> an existing CONFIG_ variable that distinguishes the two. > > > To add to this.. amd64 uses "elf_i386". Looking at i386 it uses > elf_i386_obsd and that is the only emulation supported. > > So change the list of emulations to try to > > elf_i386 elf_i386_fbsd elf_i386_obsd i386pe > > > Someone needs to take a look at what NetBSD / DragonFly are > using. My NetBSD 6.1.4 amd64 VM uses elf_i386. Building should still work on an OS which doesn't support an emulation in the list in configure -- we'll just fall back to not building the option ROMs, which is the same position any non-x86 host architecture is in. thanks -- PMM
On 8 August 2016 at 17:11, Peter Maydell <peter.maydell@linaro.org> wrote: > The various host OSes are irritatingly variable about the name > of the linker emulation we need to pass to ld's -m option to > build the i386 option ROMs. Instead of doing this via a > CONFIG ifdef, check in configure whether any of the emulation > names we know about will work and pass the right answer through > to the makefile. If we can't find one, we fall back to not trying > to build the option ROMs, in the same way we would for a non-x86 > host platform. > > This is in particular necessary to unbreak the build on OpenBSD, > since it wants a different answer to FreeBSD and we don't have > an existing CONFIG_ variable that distinguishes the two. > > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > --- > This works for Linux and for the Windows builds; I don't have any > BSD systems to test it on. Brad and Sean, can I ask you to test this > on OpenBSD and FreeBSD, please? > > (I think this is going to miss -rc2. Sorry about that; we'll > get OpenBSD builds fixed for -rc3.) Applied to master (as-is, in the absence of a build check for i386 OpenBSD; we can always add the elf_i386_obsd entry to the list as a later patch, and if not then i386 is no worse off than other non-x86-64 OpenBSD hosts: it just won't build the option roms from source). thanks -- PMM
On 8/15/2016 1:26 PM, Peter Maydell wrote: > On 8 August 2016 at 17:11, Peter Maydell <peter.maydell@linaro.org> wrote: >> The various host OSes are irritatingly variable about the name >> of the linker emulation we need to pass to ld's -m option to >> build the i386 option ROMs. Instead of doing this via a >> CONFIG ifdef, check in configure whether any of the emulation >> names we know about will work and pass the right answer through >> to the makefile. If we can't find one, we fall back to not trying >> to build the option ROMs, in the same way we would for a non-x86 >> host platform. >> >> This is in particular necessary to unbreak the build on OpenBSD, >> since it wants a different answer to FreeBSD and we don't have >> an existing CONFIG_ variable that distinguishes the two. >> >> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> >> --- >> This works for Linux and for the Windows builds; I don't have any >> BSD systems to test it on. Brad and Sean, can I ask you to test this >> on OpenBSD and FreeBSD, please? >> >> (I think this is going to miss -rc2. Sorry about that; we'll >> get OpenBSD builds fixed for -rc3.) > Applied to master (as-is, in the absence of a build check for > i386 OpenBSD; we can always add the elf_i386_obsd entry to the > list as a later patch, and if not then i386 is no worse off > than other non-x86-64 OpenBSD hosts: it just won't build the > option roms from source). > > thanks > -- PMM I am fine with that. Thank you.
diff --git a/configure b/configure index f57fcc6..7c744ad 100755 --- a/configure +++ b/configure @@ -4699,7 +4699,16 @@ roms= if test \( "$cpu" = "i386" -o "$cpu" = "x86_64" \) -a \ "$targetos" != "Darwin" -a "$targetos" != "SunOS" -a \ "$softmmu" = yes ; then - roms="optionrom" + # Different host OS linkers have different ideas about the name of the ELF + # emulation. Linux and OpenBSD use 'elf_i386'; FreeBSD uses the _fbsd + # variant; and Windows uses i386pe. + for emu in elf_i386 elf_i386_fbsd i386pe; do + if "$ld" -verbose 2>&1 | grep -q "^[[:space:]]*$emu[[:space:]]*$"; then + ld_i386_emulation="$emu" + roms="optionrom" + break + fi + done fi if test "$cpu" = "ppc64" -a "$targetos" != "Darwin" ; then roms="$roms spapr-rtas" @@ -5539,6 +5548,7 @@ fi echo "LDFLAGS=$LDFLAGS" >> $config_host_mak echo "LDFLAGS_NOPIE=$LDFLAGS_NOPIE" >> $config_host_mak echo "LD_REL_FLAGS=$LD_REL_FLAGS" >> $config_host_mak +echo "LD_I386_EMULATION=$ld_i386_emulation" >> $config_host_mak echo "LIBS+=$LIBS" >> $config_host_mak echo "LIBS_TOOLS+=$libs_tools" >> $config_host_mak echo "PTHREAD_LIB=$PTHREAD_LIB" >> $config_host_mak diff --git a/pc-bios/optionrom/Makefile b/pc-bios/optionrom/Makefile index 24e175e..5bbe233 100644 --- a/pc-bios/optionrom/Makefile +++ b/pc-bios/optionrom/Makefile @@ -41,18 +41,8 @@ build-all: multiboot.bin linuxboot.bin linuxboot_dma.bin kvmvapic.bin %.o: %.S $(call quiet-command,$(CPP) $(QEMU_INCLUDES) $(QEMU_DGFLAGS) -c -o - $< | $(AS) $(ASFLAGS) -o $@," AS $(TARGET_DIR)$@") -ifdef CONFIG_WIN32 -LD_EMULATION = i386pe -else -ifdef CONFIG_BSD -LD_EMULATION = elf_i386_fbsd -else -LD_EMULATION = elf_i386 -endif -endif - %.img: %.o - $(call quiet-command,$(LD) $(LDFLAGS_NOPIE) -m $(LD_EMULATION) -Ttext 0 -e _start -s -o $@ $<," Building $(TARGET_DIR)$@") + $(call quiet-command,$(LD) $(LDFLAGS_NOPIE) -m $(LD_I386_EMULATION) -Ttext 0 -e _start -s -o $@ $<," Building $(TARGET_DIR)$@") %.raw: %.img $(call quiet-command,$(OBJCOPY) -O binary -j .text $< $@," Building $(TARGET_DIR)$@")
The various host OSes are irritatingly variable about the name of the linker emulation we need to pass to ld's -m option to build the i386 option ROMs. Instead of doing this via a CONFIG ifdef, check in configure whether any of the emulation names we know about will work and pass the right answer through to the makefile. If we can't find one, we fall back to not trying to build the option ROMs, in the same way we would for a non-x86 host platform. This is in particular necessary to unbreak the build on OpenBSD, since it wants a different answer to FreeBSD and we don't have an existing CONFIG_ variable that distinguishes the two. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- This works for Linux and for the Windows builds; I don't have any BSD systems to test it on. Brad and Sean, can I ask you to test this on OpenBSD and FreeBSD, please? (I think this is going to miss -rc2. Sorry about that; we'll get OpenBSD builds fixed for -rc3.) configure | 12 +++++++++++- pc-bios/optionrom/Makefile | 12 +----------- 2 files changed, 12 insertions(+), 12 deletions(-)