(Thanks to ken and rsc for pointing this out)
rsc:
ken pointed out that there's a race in the new
one-lock-per-channel code. the issue is that
if one goroutine has gone to sleep doing
select {
case <-c1:
case <-c2:
}
and then two more goroutines try to send
on c1 and c2 simultaneously, the way that
the code makes sure only one wins is the
selgen field manipulation in dequeue:
// if sgp is stale, ignore it
if(sgp->selgen != sgp->g->selgen) {
//prints("INVALID PSEUDOG POINTER\n");
freesg(c, sgp);
goto loop;
}
// invalidate any others
sgp->g->selgen++;
but because the global lock is gone both
goroutines will be fiddling with sgp->g->selgen
at the same time.
This results in a 7% slowdown in the single threaded case for a
ping-pong microbenchmark.
Since the cas predominantly succeeds, adding a simple check first
didn't make any difference.
R=rsc
CC=golang-dev
https://golang.org/cl/180068
too hard to make it the same everywhere.
still checking for non-zero exit status though.
disable core files while running tests
R=r
http://go/go-review/1026037
gccgo in a few different places, and crashes 6g as of this
writing. Note that the index in a composite literal must be
looked up in a different namespace if the composite literal
type turns out to be a struct.
R=rsc
DELTA=36 (36 added, 0 deleted, 0 changed)
OCL=33681
CL=33685
let errchk exit 0 even if it has reported a BUG.
it echoed BUG and that's all that matters.
R=r
DELTA=143 (1 added, 89 deleted, 53 changed)
OCL=32533
CL=32542
4 known bugs; 0 unexpected bugs
which should be less scary to people
unfamiliar with the tests.
R=r
DELTA=44 (29 added, 12 deleted, 3 changed)
OCL=32460
CL=32464
so that golden.out does not include
the name of the compiler (which is
arch-specific and shows up in diffs).
R=r,iant
DELTA=3 (0 added, 0 deleted, 3 changed)
OCL=31980
CL=31983
introduce NodeList* type in compiler to replace OLIST.
this clarifies where lists can and cannot occur.
list append and concatenation are now cheap.
the _r rules are gone from yacc.
rev and unrev are gone.
no more lists of lists.
the representation of assignments is a bit clunkier.
split into OAS (1=1) and OAS2 (2 or more on one side).
delete dead chanrecv3 code.
delay construction of func types.
R=ken
OCL=31745
CL=31762
The change to assign.go is because the gcc testsuite fails to
handle .* in a normal way: it matches against the entire
compiler output, not just a single line.
assign.go:15:6: error: incompatible types in assignment (implicit assignment of 'sync.Mutex' hidden field 'key')
assign.go:19:6: error: incompatible types in assignment (implicit assignment of 'sync.Mutex' hidden field 'key')
assign.go:23:6: error: incompatible types in assignment (implicit assignment of 'sync.Mutex' hidden field 'key')
assign.go:27:6: error: incompatible types in assignment (implicit assignment of 'sync.Mutex' hidden field 'key')
chan/perm.go:14:5: error: incompatible types in assignment
chan/perm.go:15:5: error: incompatible types in assignment
chan/perm.go:16:6: error: incompatible types in assignment
chan/perm.go:17:6: error: incompatible types in assignment
chan/perm.go:24:7: error: invalid send on receive-only channel
chan/perm.go:25:12: error: invalid send on receive-only channel
chan/perm.go:31:4: error: invalid receive on send-only channel
chan/perm.go:32:9: error: invalid receive on send-only channel
chan/perm.go:38:2: error: invalid send on receive-only channel
chan/perm.go:42:2: error: invalid receive on send-only channel
initializerr.go:14:17: error: reference to undefined variable 'X'
initializerr.go:14:19: error: mixture of field and value initializers
initializerr.go:15:26: error: duplicate value for field 'Y'
initializerr.go:16:10: error: too many values in struct composite literal
initializerr.go:18:19: error: index expression is not integer constant
initializerr.go:17:11: error: too many elements in composite literal
R=rsc
DELTA=12 (0 added, 0 deleted, 12 changed)
OCL=29657
CL=29665
this is not a user-visible change.
before, all interface values were
struct Itype {
Sigt *type;
Sigi *inter;
void *method[n];
}
struct Iface {
void *addr;
Itype *itype;
}
the itype is basically a vtable, but it's unnecessary
if the static type is interface{ }.
for interface values with static type empty, the
new representation is
struct Eface {
void *addr;
Sigt *type;
}
this complicates the code somewhat, but
it reduces the number of Itypes that
have to be computed and cached,
it opens up opportunities to avoid function
calls in a few common cases,
and it will make it possible to lay out
interface{} values at compile time,
which i think i'll need for the new reflection.
R=ken
OCL=28701
CL=29121
