xmalloc() and xrealloc() and the mixed usage of xmalloc in some .c's from
libiberty.a and other .c's from libreadline.so produces an unusable binary
on the Alpha.
While I am here, preventatively move other libs in the link order.
Submitted by: gallatin
Correct backtrace was made more complex when the new signal trampoline
was introduced to support more than 32 signals, while keeping a modified
version of the old signal trampoline.
The 'where' command will now show:
#2 <signal handler called>
where appropiate.
Submitted by: Tor.Egge@fast.no
Rev 1.2 changed the default emulation from ``elf64_sparc'' to ``elf32_sparc''
and I never noticed it after my review of rev 1.1. Backing the change of
the default emulation out, and Wa-la!, I can now build a native [and usable]
binutils. WTF, the "-m elf64_sparc" parameter handed to `ld' by `gcc'
wasn't DTRT is beyond me.
Presumably the issue was with arparse.[ch]. Those are now in FREEBSD-Xlist
and FREEBSD-deletelist. So we do not import the Bison produced files that
was causing the problem.
Submitted by: ru
(the two may be different (ie, build vs. runtime))
Allow ldscript's SEARCH_DIR do be rooted somewhere other than `/'.
(in this case at TOOLS_PREFIX)
These changes are most helpful during `make buildworld' so that the shared
libs built in the middle of `make buildworld' are used vs. the ones in
/usr/lib on the build machine.
Submitted by: ru
code in ipl.s and icu_ipl.s that used them was removed when the
interrupt thread system was committed. Debuggers also knew about
Xresume* because these labels hide the real names of the interrupt
handlers (Xintr*), and debuggers need to special-case interrupt
handlers to get the interrupt frame.
Both gdb and ddb will now use the Xintr* and Xfastintr* symbols to
detect interrupt frames. Fast interrupt frames were never identified
correctly before, so this fixes the problem of the running stack
frame getting lost in a ddb or gdb trace generated from a fast
interrupt - e.g. when debugging a simple infinite loop in the kernel
using a serial console, the frame containing the loop would never
appear in a gdb or ddb trace.
Reviewed by: jhb, bde
are linking against does not have basename(). There is a buffer overflow
bug in lib/libc/gen/basename.c rev 1.1. There is no way for us to test
what revision of basename() we have in libc, thus this change.
Requested by: ru
misuse of /usr/src/include headers. This REALLY fixes
the 20010919 src/UPDATING entry.
With this patch the 4.2-RELEASE box was able to survive
the 5.0-CURRENT "make world".
Beat over the head with this patch: obrien
The version of the kernel has no bearing on what is in libc.
We now search for basename in libc to determin if we need to include
the libiberty version in the build.
This is all still a bit bogus as it will (like the sysctl method) cause
basename.o to be linked into the cross-build as well as the host build. It
would probably be better to test if we were doing the initial host build and
unconditionally include that. Once we've generated the target libc we know
that basename is available. (maybe test for $TOOLS_PREFIX or something).
Submitted by: peter
Note ALL MODULES MUST BE RECOMPILED
make the kernel aware that there are smaller units of scheduling than the
process. (but only allow one thread per process at this time).
This is functionally equivalent to teh previousl -current except
that there is a thread associated with each process.
Sorry john! (your next MFC will be a doosie!)
Reviewed by: peter@freebsd.org, dillon@freebsd.org
X-MFC after: ha ha ha ha
reading old a.out core files, which are totally 100% non-understandable
to the gdb floating-point reader if you have SSE turned on.
This should be the last of the world build breakers...
end of the include searching. We really need a real fix for the issue of
which set of headers to use in compiling the cross-tools -- /usr/include,
or /usr/src/include.
call and trap entry points so they're easy to find and change
- Use the cpuhead and allcpu list to locate globaldata for the current
cpu, rather than SMP_prvspace or __globaldata
- Use offsets into struct globaldata directly to find per-cpu variables,
rather than symbols in globals.o
Glanced at by: peter
when using gdb on a remote target. The fix is to restrict PT_GETDBREGS
calls to `child' and `freebsd-uthreads' targets solely.
I've been in some conversation with Brian about this, and this solution
seems to be the most appropriate one.
PR: gnu/21685
Submitted by: bsd
`wait.h' that was in contrib/binutils/, however this wait.h went away with
bintuils 2.10.0 so I `cvs rm'ed it. Now we find gdb will not build. This
binutils wait.h contained nothing we didn't already have in <sys/wait.h>.
So just hack a symlink to it.
with Brian's kernel support for i386 debug registers. This makes
watchpoints actually usable for real-life problems. Note: you can
only set watchpoints on 1-, 2- or 4-byte locations, gdb automatically
falls back to [sloooow] software watchpoints when attempting to use
them on variables which don't fit into this category. To circumvent
this, one can use the following hack:
watch *(int *)0x<some address>
David O'Brien is IMHO considering to get this fully integrated into the
official GDB, but as long as we've got the i386/* files sitting around
in our private FreeBSD tree here, the feature can now be tested more
extensively, so i'm committing this for the time being.
This work has been done in order to debug a tix toolkit problem, thus
it has been sponsored by teh Deutsche Post AG.
Reviewed by: bsd (not the operating system, but Brian :-)
libraries in LDADD so that `make checkdpadd' doesn't report non-errors.
Fixed some style bugs (the usual ones for DPADD and LDADD, and misformatting
of $FreeBSD$).
The target machine is represented by TARGET_ARCH. MACHINE_ARCH always
represents the host machine. When TARGET_ARCH is not defined, it is
assumed to be equal to MACHINE_ARCH. This means that we're building a
native toolset by default. We're creating cross-compilation tools when
MACHINE_ARCH != TARGET_ARCH.
TARGET_ARCH is defined when building binutils as part of the bootstrap
build and is set to reflect the architecture we're currently cross-
building. With this change binutils is ready for cross-building.
All Makefiles now use MACHINE_ARCH for the target architecture.
Unification is required for cross-building.
Tags added to:
sys/boot/Makefile
sys/boot/arc/loader/Makefile
sys/kern/Makefile
usr.bin/cpp/Makefile
usr.bin/gcore/Makefile
usr.bin/truss/Makefile
usr.bin/gcore/Makefile:
fixed typo: MACHINDE -> MACHINE_ARCH
tidy up the logic that works out which sub-directories to build.
The new directories with freebsdelf suffixes now have freebsd suffixes
after a repo move by Peter at the request of David O'Brien.
directory to /usr/cross/${MACHINE_ARCH}-freebsdelf/usr/lib so that
the cross tools behave the same way that the host versions do. When
building cross tools, Cygnus doesn't set the default library directory.
This doesn't suit FreeBSD IMHO.
Add WinNT emulation support too. You only get this if you've set
BINUTILSDISTDIR because the contrib/binutils repository doesn't
contain the required sources.
directory to /usr/cross/${MACHINE_ARCH}-freebsdelf/usr/lib so that
the cross tools behave the same way that the host versions do. When
building cross tools, Cygnus doesn't set the default library directory.
This doesn't suit FreeBSD IMHO.
gas for i386 targeted to NT for those (like me) who have to do work
targeted to NT, but can't stand actually looking at it all day long.
I cross build apps on FreeBSD and just run them on NT later. Life is
better that way.
Allow for the case where the host architecture might also be listed
in CROSS_ARCH, so don't do things twice. This situation can arise if you
want NT support in binutils (CROSS_ARCH=i386 CROSS_FORMAT=winnt).
* Update build for gdbserver and gdbreplay to work under binutils
* Fix gdbserver to use PT_GETREGS etc to access registers, removing the
dependancy on the u-area.
* Make gdbserver work on the alpha.
zero when building for little endian machines.
Correct the target names for mips. We just use the generic targets
for mips elf, so the mipse[lb]-unknown-freebsd emulation types don't
exist.
name of the bfd target, not the gnu-standard target name. Corrected
to be elf32-{big,little}mips from mipse[bl]-unknown-freebsd.
DEFAULT_EMULATION was bogusly defined, causing ld to always fail (this
was masked by the TARGET bogosity). Define correctly as elf32bmip and
elf32lmip. Mips doesn't follow the same conventions as i386 and alpha
do in this area.
ld now appears to work correctly for the uncommitted mips changes to
egcs.
Unlike the unisex architecutres we've had so far, mips is bisexual.
These tools can produce either byte sex, and the compiler/make
determines the proper gender to use. Otherwise, we'd have to have had
mipsel and mipseb in all the places that we have just mips. And there
are other complications with doing that (binutils doesn't like to
build mips tools without both byte genders, it seems).
Introduced BINUTIL_ARCH so that other bisexual architectures can a
generic mechanism.
We cannot just define MACHINE_ARCH as mips because we need to
differentiate big and little endian types of binaries. Discussions on
freebsd-arch have hashed out this issue (and the parallel libc
issues). NetBSD is moving towards mipsel and mipseb for their two
flavors of mips ports (in time for 1.4, if this change hasn't already
been accomplished).
I've been building i386 worlds with this tree for a three months with
these files in place with no ill effects.
can be used to select them. The purpose of this is not necessarily to
allow another host format, but to allow us to use the objformat trickery
for cross compilation.
was already put in SRCS by the general yacc rules. Putting ldemul-list.h
in SRCS fixes races in `make -jN' when .depend hasn't been made.
Don't forget to clean ldemul-list.h.
kernel's) curproc is null. This fixes endless recursion in
xfer_umem() for attempts to read from user addresses, in particular
for attempts to read %fs and %gs from the pcb for `info reg'.
gas for each target format. So for m68k targets that means several
gases. I wanted a m68k gas for VxWorks which uses aout in sun3 big
endian format, cross compiled on i386 under FreeBSD using libraries
supplied by DEC and intended by them for cross compilation on Alpha
under OSF/1. And it actually works!
support more than one architecture at a time, build as from the
default for the host and if CROSS_TOOLS defines other architectures,
build them as as_${arch}
are now included according to the cross-architecture support required,
default the BINDIR for i386 to /usr/libexec/elf here instead of in
all the i386 specific makefiles. For all other architectures, BINDIR
is just /usr/bin.
be defined (in /etc/make.conf, say) and set to the additional architectures
that need to be compiled in. So on alpha I set CROSS_TOOLS = i386.
On i386 you can't build alpha due to lack of 64-bit support on 32-bit
architectures, but that's a GNU problem.
This change relies on makefiles in the binutils sub-directories having
the extension defined in the CROSS_TOOLS, instead of those makefiles
being selected based on the host architecture.
worked because .ORDER prevented problems from concurrent generation
of multiple parsers (and their headers), and there were no missing
dependencies because the generated headers were not actually used.
requires the new file.
Fixed stale near-copy of contrib/libreadline/doc/hsuser.texinfo. Patch
it at build ntime, and only keep the patch for it here.
Don't keep a copy of contrib/gdb/gdb/doc/all-cfg.texi here. Link to it
at build time.
be worth much effort. Install all i386 binutils programs in
"/usr/libexec/elf". Disable a.out support in libbfd. It's too
dangerous to leave it in. Some of the utilities think they can
handle a.out, but they generate bad object files.
the binutils headers for (machine)-unknown-freebsdelf.
Also copy the bfd.h header to an architecture specific directory
because there are two fundamental lines that differ (32-bit vs 64-bit)
between i386 and alpha.
The config.h for libbinutils generates the same on alpha as i386,
so I didn't change that (though I was tempted!).
support building it for variant architectures. It was already
becoming clear that the former structure was too rigid and didn't
scale well.
The usual sort of makefile magic arranges to .include an architecture
specific makefile "Makefile.${MACHINE_ARCH}" in each directory
where it exists. Also, sources will be found in each subdirectory
"${MACHINE_ARCH}" that exists. This is all taken care of automatically
by the top level "Makefile.inc0".
This all seems to work right for the i386 now. I have also converted
those alpha pieces already present to the new schema as best I
could.
Also: change the BINDIR on the i386 to /usr/libexec/elf for "ar"
and "ranlib". They are not object format independent enough to
put into /usr/bin.
strip program (via "install") to strip itself. But the program
wasn't executable because "install" hadn't made it so yet. I
borrowed the method used for the old strip to get around this.
This finishes up the binutils import. But I am leaving it disabled
in "src/gnu/usr.bin/Makefile" for now. It is not used by anything
yet, so I'll take this opportunity to run one more round of tests
before enabling it.
0xefbfe000) and kernel_start (normally 0xf0100000).
Things are unnecessarily (?) difficult because procfs is used to
access user addresses in the live-kernel case although we must have
access to /dev/mem to work at all, and whatever works for the
dead-kernel case should work in all cases (modulo volatility of
live kernel variables). We used the wrong range [0, kernel_start)
for user addresses. Procfs should only work up to VM_MAXUSER_ADDRESS,
but it bogusly works for reads up to the address 2 pages higher
(the user area, including the kernel stack, is mapped to where the
user area used to be (WTUAUTB)). Procfs can not work at all for
addresses between WTUAUTB and kernel_start.
Now we use procfs only to access addresses up to VM_MAXUSER_ADDRESS.
Higher addresses are translated normally using kvtophys(), so the
user ptd is used for addresses below the real kernel start (0xf0000000;
see INKERNEL()) and nothing is found WTUAUTB.
Strange accesses that cross the user-kernel boundary are now handled,
but such ranges are currently always errors because they necessarily
overlap the hole WTUAUTB.
Short reads are still not handled.
Correct translations would have been null. However, kstack was
the top of the kernel stack instead of the base of the kernel stack
like it was when the kernel exported it, so the area above the
kernel stack was mistranslated and the kernel stack was not
translated. This bug was depended on to compensate for the wrong
value of kstack - to read the pcb, instead of just using the address
of the pcb, we used the mistranslated address of kstack, which
happened to be the same (curpcb = kstack - 0x2000).
This area is simpler than it used to be now that the kernel stack
address is per-process. The code still seems to be more complicated
than necessary - the `found_pcb == 0' case seems to be unused.
gdb was cloned from the buggy version of kvm_uread() in libkvm and
had the same bugs. It looped endlessly on EOF and checked errno
without setting it in the lseek() error check. The first bug caused
gdb to loop endlessly for reads from addresses between the end of
the user area and the start of the kernel text. kvm_uread() should
not be used for addresses beyond the end of the user area, but is
due to bugs elsewhere.
the previous frame is in the usual place even for traps, interrupts
and syscalls in the kernel, because the assembly language stubs
don't change the frame pointer. The previous frame is just not for
the calling function. We may as well depend on this as on magic to
determine the trap frame address. The magic is in FRAME_SAVED_PC()
which elides the correct number of stubs (1) to go back to a pc that
matches the previous frame.
Removing fbsd_kern_frame_chain() fixes bugs in it. Xsyscall was
misspelled as _Xsyscall (gdb removes one leading underscore), so
the tf_syscall frame type was never found. This was harmless
because tf_normal works in all cases in fbsd_kern_frame_chain()
and Xsyscall is spelled correctly in fbsd_kern_frame_saved_pc()
where it matters. There were style bugs on almost every line,
starting with a primary indent of 7.
following "panic:" or "Fatal trap". `panicstr' is still printed,
although it is redundant if there is a valid message buffer and
incomplete if it contains `%'s. I think the awk command belongs
here and not in a script since a standard format with complete
messages is good for bug reports.
Implemented reading of %fs and %gs from core files.
Print weird floating point values better. We have to convert long
doubles to doubles here because of limitations and bugs in printf()
and floatformat_to_double() (long doubles aren't really supported
and naive converion to double causes exceptions). Conversion loses
information about weird formats (everything becomes a quiet NaN),
and printf() doesn't know about different types of NaNs anyway.
