A bunch of places are a bit too picky about the architecture.
Simplify them.
Also use a large PEBASE for 64-bit systems.
This more closely matches what is usually used on Windows x86-64
and is required for Windows arm64.
Unfortunately, we still need a special case for x86-64 because
of some cgo relocations. This may be fixable separately.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I65212d28ad4d8c40e2b70dc29f7fce072babecb5
Reviewed-on: https://go-review.googlesource.com/c/go/+/288816
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
The files being deleted contain no code.
They exist because back when we used Makefiles
that listed all the Go sources to be compiled, we wrote
patterns like syscall_$GOOS_$GOARCH.go,
and it was easier to create dummy empty files
than introduce conditionals to not look for that
file on Windows.
Now that we have the go command instead,
we don't need those dummy files.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: Ie0066d1dd2bf09802c74c6a496276e8c593e4bc2
Reviewed-on: https://go-review.googlesource.com/c/go/+/288815
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Frame pointers were already enabled on linux, darwin, ios,
but not freebsd, android, openbsd, netbsd.
But the space was reserved on all platforms, leading to
two different arm64 framepointer conditions in different
parts of the code, one of which had no name
(framepointer_enabled || GOARCH == "arm64",
which might have been "framepointer_space_reserved").
So on the disabled systems, the stack layouts were still
set up for frame pointers and the only difference was not
actually maintaining the FP register in the generated code.
Reduce complexity by just enabling the frame pointer
completely on all the arm64 systems.
This commit passes on freebsd, android, netbsd.
I have not been able to try it on openbsd.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I83bd23369d24b76db4c6a648fa74f6917819a093
Reviewed-on: https://go-review.googlesource.com/c/go/+/288814
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Document the various hard-coded architecture checks
or remove them in favor of more general checks.
This should be a no-op now but will make the arm64 port
have fewer diffs.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: Ifd6b19e44e8c9ca4a0d2590f314928ce235821b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/288813
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Externalthreadhandler was not handling its own stack correctly.
It incorrectly referred to the saved LR slot (uninitialized, it turned out)
as holding the return value from the called function.
Externalthreadhandler is used to call two different functions:
profileloop1 and ctrlhandler1.
Profileloop1 does not return, so no harm done.
Ctrlhandler1 returns a boolean indicating whether the handler
took care of the control event (if true, no other handlers run).
It's hard to say exactly what uninitialized values are likely to
have been returned instead of ctrlhandler1's result, but it
probably wasn't helping matters.
Change-Id: Ia02f1c033df618cb82c2193b3a8241ed048a8b18
Reviewed-on: https://go-review.googlesource.com/c/go/+/288812
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
This code has clearly never run successfully,
since one of the “tail calls" calls the wrong function,
and both of them appear in functions with stack frames
that are never going to be properly unwound.
Probably there is no windows/arm under WINE at all.
But might as well fix the code.
Change-Id: I5fa62274b3661bc6bce098657b5bcf11d59655eb
Reviewed-on: https://go-review.googlesource.com/c/go/+/288811
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
The isAbort check was wrong for non-x86 systems.
That was causing the exception chain to be passed back to Windows.
That was causing some other kind of fault - not sure what.
That was leading back to lastcontinuehandler to print a larger
stack trace, and then the throwing = 1 print added runtime.abort,
which made TestAbort pass even though it wasn't really working.
Recognize abort properly and handle it as Go, not as something
for Windows to try to handle.
Keep the throwing = 1 print, because more detail on throw is
always better.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: If614f4ab2884bd90410d29e28311bf969ceeac09
Reviewed-on: https://go-review.googlesource.com/c/go/+/288810
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
If traceback fails, it prints a helpful hex dump of the stack.
But the hex numbers have no 0x prefix, which might make it
a little unclear that they are hex.
We only print two per line, so there is plenty of room for the 0x.
Print it, which lets us delete a custom hex formatter.
Also, in the translated <name+off> hints, print off in hex
(with a 0x prefix). The offsets were previously decimal, which
could have been confused for hex since none of the hex had
0x prefixes. And decimal is kind of useless anyway since the
offsets shown in the main traceback are hex, so you can't
easily match them up without mental base conversions.
Just print hex everywhere, clearly marked by 0x.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I72d26a4e41ada38b620bf8fe3576d787a2e59b47
Reviewed-on: https://go-review.googlesource.com/c/go/+/288809
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
The constant was wrong, and the “right” constant doesn't work either.
But with the actually-right constant (and possibly earlier fixes in this
stack as well), profiling now works.
Change-Id: If8caff1da556826db40961fb9bcfe2b1f31ea9f9
Reviewed-on: https://go-review.googlesource.com/c/go/+/288808
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Also give up on the fiction that these files can be regenerated.
They contain many manual edits, and they're fairly small anyway.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: Ib4e4e20a43d8beb1d5390fd184160c33607641f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/288807
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
The architecture-specific interpretation of m->tls[0]
is unnecessary and fragile. Delete it.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I927345e52fa2f1741d4914478a29d1fb8acb0dc3
Reviewed-on: https://go-review.googlesource.com/c/go/+/288806
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
This was added in 2018 to fix a runtime crash during unwind
during a unhandled-panic-induced crash.
(See https://golang.org/cl/90895 and #23576.)
Clearly we cannot unwind past this function, and the change
did stop the unwind. But it's not a top-of-stack function, and
the real issue is that SP is changed.
The new SPWRITE bit takes care of this instead, so we can drop
it from the topofstack function.
At this point the topofstack function is only checking the
TOPFRAME bit, so we can inline that into the one call site.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I856552298032770e48e06c95a20823a1dbd5e38c
Reviewed-on: https://go-review.googlesource.com/c/go/+/288805
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
I added morestack to this list in 2013 with an explanation
that they were needed if we “start a garbage collection on g0
during a stack split or unsplit”.
(https://golang.org/cl/11533043)
This explanation no longer applies for a handful of reasons,
most importantly that if we did stop a stack scan in the middle
of a call to morestack, we'd ignore pointers above the split,
which would lead to memory corruption. But we don't scan
goroutine stacks during morestack now, so that can't happen.
If we did see morestack during a GC, that would be a good time
to crash the program.
The real problem with morestack is during profiling, as noted
in the code review conversation during 2013. And in profiling
we just need to know to stop and not unwind further, which
the new SPWRITE bit will do for us.
So remove from topofstack and let the program crash if GC
sees morestack and otherwise let the SPWRITE stop morestack
unwinding during profiling.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I06d95920b18c599c7c46f64c21028104978215d7
Reviewed-on: https://go-review.googlesource.com/c/go/+/288804
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
I added mcall to this list in 2013 without explaining why.
(https://codereview.appspot.com/11085043/diff/61001/src/pkg/runtime/traceback_x86.c)
I suspect I was stopping crashes during profiling where the unwind
tried to walk up past mcall and got confused.
mcall is not something you can unwind past, because it switches
stacks, but it's also not something you should expect as a
standard top-of-stack frame. So if you do see it during say
a garbage collection stack walk, it would be important to crash
instead of silently stopping the walk prematurely.
This CL removes it from the topofstack list to avoid the silent stop.
Now that mcall is detected as SPWRITE, that will stop the
unwind (with a crash if encountered during GC, which we want).
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I666487ce24efd72292f2bc3eac7fe0477e16bddd
Reviewed-on: https://go-review.googlesource.com/c/go/+/288803
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
No change to actual runtime, but helps reduce the laundry list
of functions.
mcall, morestack, and asmcgocall are not actually top-of-frame,
so those need more attention in follow-up CLs.
mstart moved to assembly so that it can be marked TOPFRAME.
Since TOPFRAME also tells DWARF consumers not to unwind
this way, this change should also improve debuggers a
marginal amount.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: If1e0d46ca973de5e46b62948d076f675f285b5d9
Reviewed-on: https://go-review.googlesource.com/c/go/+/288802
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
The old code was very clever about predicting whether a traceback was safe.
That cleverness has not aged well. In particular, the setsSP function is missing
a bunch of functions that write to SP and will confuse traceback.
And one such function - jmpdefer - was handled as a special case in
gentraceback instead of simply listing it in setsSP.
Throw away all the clever prediction about whether traceback will crash.
Instead, make traceback NOT crash, by checking whether the function
being walked writes to SP.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I3d55fe257a22745e4919ac4dc9a9378c984ba0da
Reviewed-on: https://go-review.googlesource.com/c/go/+/288801
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
The runtime traceback code has its own definition of which functions
mark the top frame of a stack, separate from the TOPFRAME bits that
exist in the assembly and are passed along in DWARF information.
It's error-prone and redundant to have two different sources of truth.
This CL provides the actual TOPFRAME bits to the runtime, so that
the runtime can use those bits instead of reinventing its own category.
This CL also adds a new bit, SPWRITE, which marks functions that
write directly to SP (anything but adding and subtracting constants).
Such functions must stop a traceback, because the traceback has no
way to rederive the SP on entry. Again, the runtime has its own definition
which is mostly correct, but also missing some functions. During ordinary
goroutine context switches, such functions do not appear on the stack,
so the incompleteness in the runtime usually doesn't matter.
But profiling signals can arrive at any moment, and the runtime may
crash during traceback if it attempts to unwind an SP-writing frame
and gets out-of-sync with the actual stack. The runtime contains code
to try to detect likely candidates but again it is incomplete.
Deriving the SPWRITE bit automatically from the actual assembly code
provides the complete truth, and passing it to the runtime lets the
runtime use it.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I227f53b23ac5b3dabfcc5e8ee3f00df4e113cf58
Reviewed-on: https://go-review.googlesource.com/c/go/+/288800
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
A g's sched.g is set in newproc1:
newg.sched.g = guintptr(unsafe.Pointer(newg))
After that, it never changes. Yet lots of assembly code does
"g.sched.g = g" unnecessarily. Remove all those lines to avoid
confusion about whether it ever changes.
Also, split gogo into two functions, one that does the nil g check
and a second that does the actual switch. This way, if the nil g check
fails, we get a stack trace showing the call stack that led to the failure.
(The SP write would otherwise cause the stack trace to abort.)
Also restore the proper nil g check in a handful of assembly functions.
(There is little point in checking for nil g *after* installing it as the real g.)
Change-Id: I22866b093f901f765de1d074e36eeec10366abfb
Reviewed-on: https://go-review.googlesource.com/c/go/+/292109
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
It was being rejected. Now it isn't and can be used in the runtime.
Change-Id: I4626bf9fc2e0bc26fffb87d11bede459964324b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/292129
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Both asmcgocall and systemstack need to save the calling Go code's
context for use by traceback, but they do it differently.
Systemstack's appraoch is better, because it doesn't require a
special case in traceback.
So make them both use that.
While we are here, the fake mstart caller in systemstack is
no longer needed and can be removed.
(traceback knows to stop in systemstack because of the writes to SP.)
Also remove the fake mstarts in sys_windows_*.s.
And while we are there, fix the control flow guard code in sys_windows_arm.s.
The current code is using pointers to a stack frame that technically is gone
once we hit the RET instruction. Clearly it's working OK, but better not to depend
on data below SP being preserved, even for just a few instructions.
Store the value we need in other registers instead.
(This code is only used for pushing a sigpanic call, which does not
actually return to the site of the fault and therefore doesn't need to
preserve any of the registers.)
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: Id1e3ef5e54f7ad786e4b87043f2626eba7c3bbd9
Reviewed-on: https://go-review.googlesource.com/c/go/+/288799
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Large enum sets should be sorted by name when the
values don't matter, as they don't here. Do that.
Also replace the large switch with a map lookup.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: Ibe727b5d8866bf4c40c96020e1f4632bde7efd59
Reviewed-on: https://go-review.googlesource.com/c/go/+/288798
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This can happen on Windows when recording profile samples for system threads.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I5a7ba32b1900a69f3b7acada9cb6cf8396d8a03f
Reviewed-on: https://go-review.googlesource.com/c/go/+/288797
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
The default GOROOT has nothing to do with system details.
Move it next to its one use in package runtime.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I1a601fad6335336b4616b834bb21bd8437ee1313
Reviewed-on: https://go-review.googlesource.com/c/go/+/288796
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Document what the values in internal/sys mean.
Remove various special cases for arm64 in the code using StackAlign.
Delete Uintreg - it was for GOARCH=amd64p32,
which was specific to GOOS=nacl and has been retired.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I40e8fa07b4e192298b6536b98a72a751951a4383
Reviewed-on: https://go-review.googlesource.com/c/go/+/288795
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This is dead code and need not be ported to each architecture.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I2d0072b377f73e49d7158ea304670c26f5486c59
Reviewed-on: https://go-review.googlesource.com/c/go/+/288794
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
During a cgocallback, the runtime calls needm to get an m.
The calls made during needm cannot themselves assume that
there is an m or a g (which is attached to the m).
In the old days of making direct system calls, the only thing
you had to do for such functions was mark them //go:nosplit,
to avoid the use of g in the stack split prologue.
But now, on operating systems that make system calls through
shared libraries and use code that saves state in the g or m
before doing so, it's not safe to assume g exists. In fact, it is
not even safe to call getg(), because it might fault deferencing
the TLS storage to find the g pointer (that storage may not be
initialized yet, at least on Windows, and perhaps on other systems
in the future).
The specific routines that are problematic are usleep and osyield,
which are called during lock contention in lockextra, called
from needm.
All this is rather subtle and hidden, so in addition to fixing the
problem on Windows, this CL makes the fact of not running on
a g much clearer by introducing variants usleep_no_g and
osyield_no_g whose names should make clear that there is no g.
And then we can remove the various sketchy getg() == nil checks
in the existing routines.
As part of this cleanup, this CL also deletes onosstack on Windows.
onosstack is from back when the runtime was implemented in C.
It predates systemstack but does essentially the same thing.
Instead of having two different copies of this code, we can use
systemstack consistently. This way we need not port onosstack
to each architecture.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: I3352de1fd0a3c26267c6e209063e6e86abd26187
Reviewed-on: https://go-review.googlesource.com/c/go/+/288793
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
We print things like “exit status 3221225477”
but the standard Windows form is 0xc0000005.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.
Change-Id: Iefe447d4d1781b53bef9619f68d386f2866b2934
Reviewed-on: https://go-review.googlesource.com/c/go/+/288792
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
mvs.Downgrade is pretty clearly intended to match Algorithm 4 from the
MVS blog post (https://research.swtch.com/vgo-mvs#algorithm_4).
Per the blog post:
“Downgrading one module may require downgrading other modules, but we
want to downgrade as few other modules as possible. … To avoid an
unnecessary downgrade to E 1.1, we must also add a new requirement on
E 1.2. We can apply Algorithm R to find the minimal set of new
requirements to write to go.mod.”
mvs.Downgrade does not match that behavior today: it fails to retain
the selected versions of transitive dependencies that are not implied
by downgraded direct dependencies of the target (module E in the
post). This bug is currently masked by the fact that we only call
Downgrade today with a *modload.mvsReqs, for which the Required method
happens to return the complete build list — rather than only the
direct dependencies as documented for the mvs.Reqs interface.
For #36460
Change-Id: If9c8f413b156b5f67c02787d9359394e169951b0
Reviewed-on: https://go-review.googlesource.com/c/go/+/287633
Trust: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
For #36460
Change-Id: I4f8bf0fb8dfa508b346acb3868302452409ee9da
Reviewed-on: https://go-review.googlesource.com/c/go/+/289696
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
For #36460
Change-Id: I1620c23819263ef82e571fc4d4c778277842c02d
Reviewed-on: https://go-review.googlesource.com/c/go/+/288535
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
For #36460
Change-Id: I5a8be8f36fb8825ffa08ed1427cb1e15b106b31a
Reviewed-on: https://go-review.googlesource.com/c/go/+/287732
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Req is supposed to return “a minimal requirement list”
that includes each of the module paths listed in base.
Currently, if base contains duplicates Req emits duplicates,
and a list containing duplicates is certainly not minimal.
That, in turn, requires callers to be careful to deduplicate the base
slice, and there are multiple callers that are already quite
complicated to reason about even without the added complication of
deduplicating slices.
For #36460
Change-Id: I391a1dc0641fe1dd424c16b7a1082da0d00c7292
Reviewed-on: https://go-review.googlesource.com/c/go/+/287632
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
GOARM=6 executables fail to launch on windows/arm, so set this to ARMv7
like we do for Android.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
Change-Id: Ifa13685e7ab6edd367f3dfec10296e376319dbd4
Reviewed-on: https://go-review.googlesource.com/c/go/+/291629
Reviewed-by: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org>
Unbreak the linux/riscv64 port by storing the zero value register to memory,
rather than the current code that is moving a zero intermediate to the stack
pointer register (ideally this should be caught by the assembler). This was
broken in CL#272568.
On riscv64 a zero immediate value cannot be moved directly to memory, rather
a register needs to be loaded with zero and then stored. Alternatively, the
the zero value register (aka X0) can be used directly.
Change-Id: Id57121541d50c9993cec5c2270b638b184ab9bc1
Reviewed-on: https://go-review.googlesource.com/c/go/+/292894
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Fixes#44340
Change-Id: Id80dd1f44a988b653933732afcc8e49a826affc4
Reviewed-on: https://go-review.googlesource.com/c/go/+/293209
Reviewed-by: Andrew G. Morgan <agm@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
This reverts commit CL 289712 (afd67f3). It breaks x/tools tests, and
those tests highlight that perhaps I didn't think through the
repercussions of this change as much as I should have.
Fixes#44316
Change-Id: I5db39b4e2a3714131aa22423abfe0f34a0376192
Reviewed-on: https://go-review.googlesource.com/c/go/+/292751
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Turns out that file is not formatted properly in the dev.regabi
branch.
Change-Id: I93125e65d5d3e8448c6ec1f077332c9bf7f0dd26
Reviewed-on: https://go-review.googlesource.com/c/go/+/292594
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
This CL merges the dev.regabi branch to the master branch.
In the dev.regabi branch we have refactored the compiler, and laid
some preliminary work for enabling a register-based ABI (issue #40724),
including improved late call/return lowering, improved ABI wrapper
generation, reflect call prepared for the new ABI, and reserving
special registers in the internal ABI. The actual register-based ABI
has not been enabled for the moment. The ABI-related changes are behind
GOEXPERIMENT=regabi and currently off by default.
Updates #40724, #44222.
Fixes#44224.
Change-Id: Id5de9f734d14099267ab717167aaaeef31fdba70
Currently, we call Warnl in SSA backend when we see a function
(defined or called) with regparams pragma. Calling Warnl in
concurrent environment is racy. As the debugging output is
temporary, for testing purposes we just pass -c=1. We'll remove
the pragma and the debugging print some time soon.
Change-Id: I6f925a665b953259453fc458490c5ff91f67c91a
Reviewed-on: https://go-review.googlesource.com/c/go/+/291710
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
The internal/abi package is used by runtime and needs to be
copied.
Fix longtest builders.
Change-Id: I7a962df3db2c6bf68cc6a7da74b579f381920009
Reviewed-on: https://go-review.googlesource.com/c/go/+/292592
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
(The corresponding update for the last release cycle was CL 248038.)
For #40705.
Change-Id: I13becdc4c3718a1c6986876ec56879cce3bcb34f
Reviewed-on: https://go-review.googlesource.com/c/go/+/275297
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
This is an exact port of CL 290911 to go/types.
For #31793
Change-Id: I28c42727735f467a5984594b455ca58ab3375591
Reviewed-on: https://go-review.googlesource.com/c/go/+/291319
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
This is a port of CL 289570 to go/types. It has some notable differences
with that CL:
+ A new _BadDecl error code is added, to indicate declarations with bad
syntax.
+ declInfo is updated hold not an 'alias' bool, but an aliasPos
token.Pos to identify the location of the type aliasing '=' token.
This allows for error messages to be accurately placed on the '='
For #31793
Change-Id: Ib15969f9cd5be30228b7a4c6406f978d6fc58018
Reviewed-on: https://go-review.googlesource.com/c/go/+/291318
Trust: Robert Findley <rfindley@google.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Currently these two functions assume that constants in internal/abi are
set correctly, but we actually just made them zero if
GOEXPERIMENT_REGABI is set. This means reflectcall is broken. Fix it by
stubbing out these routines even if GOEXPERIMENT_REGABI is set.
Change-Id: I4c8df6d6af28562c5bb7b85f48c03d37daa9ee0d
Reviewed-on: https://go-review.googlesource.com/c/go/+/292650
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
This change sets the register count constants to zero for the
GOEXPERIMENT regabi because currently the users of it (i.e. reflect)
will be broken, since they expect Go functions that implement the new
ABI.
Change-Id: Id3e874c61821a36605eb4e1cccdee36a2759f303
Reviewed-on: https://go-review.googlesource.com/c/go/+/292649
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
Fixes breakage accidentally introduced by https://golang.org/cl/291711.
Fixes#44295
Change-Id: I76f3e5577d1d24027d4ed2a725b5b749ab2d059c
Reviewed-on: https://go-review.googlesource.com/c/go/+/292629
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The runtime imports the internal/abi package. Recognize
internal/abi as a runtime dependent, to make trampoline generation
algorithm work.
Fix ARM build.
Change-Id: I26b6778aa41dcb959bc226ff04abe08a5a82c4f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/292610
Reviewed-by: Than McIntosh <thanm@google.com>
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Currently, in the trampoline generation pass we expect packages
are laid out in dependency order, so a cross-package jump always
has a known target address so we can check if a trampoline is
needed. With linknames, there can be cycles in the package
dependency graph, making this algorithm no longer work. For them,
as the target address is unkown we conservatively generate a
trampoline. This may generate unnecessary trampolines (if the
packages turn out laid together), but package cycles are extremely
rare so this is fine.
Updates #44073.
Change-Id: I2dc2998edacbda27d726fc79452313a21d07787a
Reviewed-on: https://go-review.googlesource.com/c/go/+/292490
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
The package documentation referenced sample metadata that was removed in CL 282632. Update this documentation to be less specific
about what metadata is available.
Additionally, the documentation on the Sample type referred to Descriptions instead of All as the source of metrics names.
Fixes#44280.
Change-Id: I24fc63a744bf498cb4cd5bda56c1599f6dd75929
Reviewed-on: https://go-review.googlesource.com/c/go/+/292309
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
Trust: Dmitri Shuralyov <dmitshur@golang.org>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
