These changes were made to etc/profile by PulkoMandy last year, but he missed
making the change in SetupEnvironment. etc/profile calls SetupEnvironment,
so we can consolidate all this there.
Also clean up some syntax.
This contains the contents of Haiku's sources, which is necessary
to include in "with source" builds for proper (L)GPL compliance,
mostly because we have GPL code in the tree.
It seems the old one was just broken (see mailing list discussion.)
This new one was built from the generic one, but it was done using
a trunk build of Clang 8 with tuned optimization flags, and is
smaller (101 instructions) than GCC 8's (134 instructions) and
the old hand-optimized one (~125 instructions?) as well as being
targeted for ARMv7 instead of ARMv6 like the old one was.
In the future, we may want to look at newlib's ARM memcpy, which
is probably much more hand-optmized than this one is, but this
is at least better than what we had before.
"cpp" is the system C preprocessor, not the one from our cross-compiler,
and in the case of my system which does not have GCC installed at all,
it doesn't even exist.
With this, Clang-ARM builds successfully create a "haiku-arm.mmc".
I couldn't get it to output, even after blessing it with "rune",
but that may just be my fault...
Now that HOST_CC is actually passed in, we need to default
everything to it; otherwise, it's up to the Jambase as to
what CC we are actually using.
Found by trying to build Haiku on a system that has no "cc"
executable, but Jam tried to use it anyway (as all three of CC,
C++, and LINK.)
We do a lot of checks against the CC specified in environment,
but we never pass it into the BuildConfig, so we have no guarantee
that's what Jam's actually using.
Additionally remove HAIKU_BOOT_BOARD while we're at it (nothing uses
it anymore), and reorder the sections in BuildConfig to make a little
more sense.
The former is passed to the compiler when linking using it,
the latter is passed to ld when it is invoked directly.
Also modify ArchitectureRules to not overwrite this setting.
GCC provides these but Clang does not; these come from the ones Clang
was going to provide, but decided against it in favor of letting libc vendors
distribute them instead, hence the license headers.
For now, they are only enabled when building with Clang, as we still use GCC's
versions otherwise. But I've tested them with apps built with GCC 7 (as Clang
builds are still too broken at present) and they seem to work just fine. (They
will not build with GCC 2 of course, due to how they use __attribute__.)
This reverts commit 033f3450e3.
The discussion on the mailing list has tended towards changing the BFont API
so that these kind of fonts return true for IsFixed() as well as IsFullAndHalfFixed
if only to avoid the API ugliness of be_fixed_font->IsFixed() returning
false.
But since that change should not really affect applications, as BeOS ones will
probably not support CJK characters at all and Haiku ones should work properly,
and the present state of affairs has been pretty annoying to CJK users, we can
at least allow users to select full-and-half-fixed fonts as their default
fixed font, and begin dealing with whatever bugs that may cause now.
The ICU class is named MessageFormat, but on Haiku, it sounds too much
like something related to BMessage (which it isn't in the slightest)
and not part of the Locale system. It works almost entirely with BStrings,
so naming it BStringFormat makes much more sense.
OK'ed by PulkoMandy and Humdinger.
This rule process the entire target's source files at once, and so
whoever wrote this rule in the first place (PulkoMandy?) probably
assumed without even testing that "cc -E" would create multiple
outputs for multiple inputs.
It doesn't, though: it just outputs them in sequence on the command line
the same way it does when the files are piped in through "cat". This
also has other advantages (e.g. preprocess errors caused by the compiler
assuming it was C not C++ code and so not defining __cplusplus, local
includes are now resolved properly, etc.)
Doing it this way does exposes other problems like the one fixed in the
previous commit (headers with no context defined, which worked previously
only because they used the context of the preceding `cat`'ed file.)
We now also remove the .pre file after collecting the catkeys.
Two files remain: "argv.c" and "argv.h", as these are still used by some
other, non-FS-related tests.
This was one of the last significant chunks of "no commercial use" code
in the tree, as it originally came from the "Filesystem Construction Kit,"
the sample code distributed with "Practical Filesystem Design with the Be FS",
and contained a sigifincant amount of BeOS kernel filesystem code in it
(the original README, not preserved here, noted that "the cache code is
the real Release 4 BeOS disk cache code".)
It has long been replaced by Ingo's from-scratch, MIT-licensed fs_shell
that lives in src/tools/fs_shell. Perhaps a note should be made somewhere
of this code's prior existence, but I don't think we will have much use for
it again, so confined to the dustbin of Git history it shall be.
src/tools is for tools used in the build or other miscellaneous utilites
that go along with Haiku itself. btrfs is, at least for now, just another
filesystem addon, so its shell should live here with all the others.
No functional change intended.
Otherwise, Clang warns that we haven't set an architecture on ARM,
which is set in CCFLAGS. Since these might also contain other flags
that affect the preprocessor, there isn't any good reason to not
pass it through, so do that.
It seems that at on some platforms at least, Clang uses @define instead
of #define, but with functionally identical syntax, so use sed to
process it as such.
Spotted while reading through the VM code while thinking about how to
implement vfork().
When axeld disabled this in 2005 (!), Haiku's kernel was still young,
BeOS drivers were still "a thing," and there was no distinction in this
function from being called by the kernel / not by the kernel.
Now, it's 2018, we manage all drivers ourselves, have SMAP enabled by
default when available, and as axeld recently noted on the mailing lists,
"there's not much reason we still use GCC2 for the kernel anyway." So we
probably don't care about any BeOS drivers that may be broken by this
(are there any still around?)
Besides the usual fixes to get this 13-year-old chunk to work again, there
are two functional changes:
1) Allow the kernel to clone whatever it likes into the user's address space.
It seems that this is often done legitimately (e.g. team creation), and
so attempting to distinguish those cases seems more work than it may be
worth right now.
The disadvantage is that drivers without proper checks may be "tricked"
into cloning areas they shouldn't; but I'm guessing if that's the case,
then something else is probably broken and the driver should be fixed.
It seems the reverse case (cloning a userland area into the kernel)
is much more common (in fact, it looks like all 4 of the 4 places
where clone_area is used in kernel-space outside the kernel itself
are doing this.)
2) At KDEBUG_LEVEL 2 and higher, throw a panic when attempting to clone
an area that does not have the protection flag set. This should make
finding any bugs exposed by this change much easier than "hardware doesn't
work" / "black screen on boot" / etc., as well as any potential future
bugs introduced in the process of driver development.
Affects *all* nodes, i.e., both directories and files. It is still possible
to override these, but it must be done in another system_package, not just by
mounting a package with newer datestamps on the conflicting files.
This is the pure version of axeld's proposal 2(d) from #10071. However,
as the haiku_secondary packages contain application mime_db entires and
are also 'system_package's, it does not fully solve that ticket.
This function is used when serializing a BPackageInfo object, which is not
done very often, and flags are used even less often. But both of these things
are done for the system packages, as they have the "system_package" flag
set, and are processed by "update_package_requires", and so this meant
that the "system_package" flag was turned into the "approve_license" flag.
This commit does not amount to any functional change: we still compare
purely by modification date alone. But it does de-duplify the code somewhat,
and makes it possible to easily implement changes to the comparison.
Untested, but this now better matches how the FreeBSD code does it.
Potentially fixes a bug where interrupts would be spuriously re-enabled
after the driver had disabled them, which may fix a KDL humdinger
reported to me via IRC.
It will probably be just stubs for the significant future, but,
here it is anyway.
Regarding the naming: Yes, the official name is "aarch64." However,
Linux, FreeBSD, and Zircon all call it "arm64", and so we will do the same.
I've configured it initially to be a Clang-only port, making no
changes to GCC buildtools whatsoever here. We'll see if that sticks,
however.
It seems not all of the kernel includes this, but some use new/delete
anyway. Further, operator delete[] was not implemented at all.
Possibly fixes the ARM build.
Since they were being Use'd at global scope, the FreeBSD versions
of a number of POSIX headers were being used instead of the POSIX ones,
which breaks the build on non-x86 as these headers presently do not work
elsewhere.
Instead, just include the base directory, and then include the compat
headers manually in the one place that actually requires them.
Also fix all the other files that expected errno.h and others to
be included implicitly, which they now are not.
This should fix the PowerPC and (part of the) ARM builds.
Previously we initialized variables and ran a few $CC tests first
(which was what the old version of --update needed.) Now, we can do it
almost immediately after the script begins.
Spotted by running --update on a GCCless system (as the $CC tests
gave warning messages due to no GCC, while CC= was set in --update
environs.)
Clang's equivalent of .tfloat is .x86_fp80, but apparently that
is not enabled for Haiku. So for now, use .double instead.
This is the same thing MinGW does.
This block was originally disabled in libbind, and we enabled it
independently; and so did NetBSD. But they also made one other
fix:
From: christos <[email protected]>
Date: Thu, 28 Dec 2017 15:12:15 +0000
Subject: [PATCH] PR/52837: Michael Kaufmann: getaddrinfo() resolves "127.0.0.1
www.example.com" to 127.0.0.1. Apply the patch from FreeBSD and explain the
rationale.
I attempted to merge NetBSD's new irs code, but their changes turned
out to be too invasive to do trivially, but in the process I had to
read our commit logs to find what we've changed, so at least I can
mark that.
Commits merged from the semi-official Git mirror of NetBSD
trunk (https://github.com/IIJ-NetBSD/netbsd-src/).
Commit authors/messages in chronological order follow:
---------------------------------------
From: christos <[email protected]>
Date: Thu, 28 Sep 2017 23:32:01 +0000
Subject: [PATCH] PR/52578: Benjamin M. Schwartz Set the AD bit when DNSSEC is
enabled (RFC 6840 Section 5.7).
From: lukem <[email protected]>
Date: Fri, 2 Mar 2018 06:31:53 +0000
Subject: [PATCH] PR/48585: Set errno when returning NULL for AF_INET
In inet_ntop4(), errno wasn't set before returning NULL.
Seems like an oversight in the existing fix for PR/48585.
Noticed by code inspection.
This requires a trunk build of Clang (the flag was only implemented
& introduced 12 days ago), but at present, full builds will fail
due to an unrelated Clang bug: https://bugs.llvm.org/show_bug.cgi?id=38356
* __NO_INLINE__ fixes the cross-build on some glibc-based systems with
newer compilers, as it prevents glibc from declaring functions inline
that we override in libroot_build.
* We can now enable tree-vrp as long as no-delete-null-pointer-checks
goes where it used to.
Only partially "tested", but I've audited this code pretty thoroughly, and
it seems the accelerant uses clone_area instead of direct pointer accesses
in all the correct places, so this should work.
This reverts commit ec1b18c58a.
This was not well enough reviewed, and it seems that at least some
consumers of the old API (e.g. WebPositive) need more than the new one provides.
Change-Id: Ie7ad1fc70dab889922424298661504b00f66d31d
This reverts commit 11dee0444f.
This reverts commit 1641a0516d.
This reverts commit 484f64f315.
These changes break various edge-cases in BTextInput (e.g. typing off the end
of a long string, using the arrow keys to navigate while out-of-bounds),
and nobody has found any fixes for them in the nearly-a-year since they've
been instanted, so they are now being reverted.
Reopens#12608, #13796.
Previously, we wrapped FreeBSD's accessors using our own in*/out* assembly
macros. Now, we skip that and just use FreeBSD's macros and assembly
directly. In the process, I've added some proper abstraction,
paving the way for the use of the FreeBSD layer on non-x86 platforms
(which will be sooner rather than later, I hope!).
This introduces some new functions required by some of the drivers
I was attempting to merge (wavelanwifi...), but as it also now
mirrors FreeBSD's bus access mechanisms exactly, it's possible
that some nuances lost before are now preserved. So this has
the potentiality to help with some of the stranger timeouts/failures
on devices that work just fine on FreeBSD.
Tested on VMware (pcnet), VirtualBox (pcnet, ipro1000), and
a T61 (iprowifi4965.)
Augustin Cavalier