x509 crypto Test fails

[Hugo Vincent]

> --- FAIL: x509.TestCertificateParse
> unhandled critical extension
> Wrong number of certs: got 0 want 2
> FAIL
> make[1]: *** [test] Error 1
> make: *** [crypto/x509.test] Error 2

[Adam Langley]

On Sat, Dec 5, 2009 at 7:57 AM, Hugo Vincent <hugo.v...@gmail.com> wrote:
>> --- FAIL: x509.TestCertificateParse
>>       unhandled critical extension
>>       Wrong number of certs: got 0 want 2

I'm afraid that I can't reproduce this on any platform.

Does anyone else see this?


AGL

[Russ Cox]

I saw this Monday or Tuesday but believed it was due to
my having broken the 8g optimizer. I reverted all my
changes (never having checked them in) and it went away.
Unfortunately, that doesn't explain this - maybe it was a
different, nondeterministic bug.

Hugo, what does "hg identify" print? (That lets us
know which version of the tree you're using.)
Also, what operating system are you running and
which compiler (6g, 8g)?

Thanks.
Russ

[Hugo Vincent]

I'm using Snow Leopard, 6g.

$ hg identify
bdfc3faa253a+ tip

Hugo

[Joshua]

I have the error too on Snow Leopard:

$ hg identify
bdfc3faa253a tip

Joshua

On Dec 5, 4:38 pm, Hugo Vincent <hugo.vinc...@gmail.com> wrote:
> On 5/12/2009, at 9:23 PM, Russ Cox wrote:
>
>
>
>
>
> > On Sat, Dec 5, 2009 at 12:09, Adam Langley <a...@golang.org> wrote:

[baldmountain]

On Dec 5, 3:09 pm, Adam Langley <a...@golang.org> wrote:
> Does anyone else see this?

Actually, I see if a different error. rpc is failing for me. It fails
like:

2009/12/05 20:35:20 Test RPC server listening on 0.0.0.0:50023
2009/12/05 20:35:20 Test HTTP RPC server listening on 0.0.0.0:50024
--- FAIL: rpc.TestHTTPRPC
Add: expected 4121272 got 15
2009/12/05 20:35:20 rpc: client protocol error: gob: type mismatch: no
fields matched compiling decoder for Reply
2009/12/05 20:35:20 rpc: client protocol error: gob: type mismatch: no
fields matched compiling decoder for Reply
2009/12/05 20:35:20 rpc: tearing down Arith connection: gob: type
mismatch: no fields matched compiling decoder for Args
2009/12/05 20:35:20 ArgNotPointer arg type not a pointer: rpc.Args
2009/12/05 20:35:20 ArgNotPointerToStruct arg type not a pointer to a
struct: *int
2009/12/05 20:35:20 ArgNotPublic argument type not public: *rpc.local
2009/12/05 20:35:20 ReplyNotPointer reply type not a pointer:
rpc.Reply
2009/12/05 20:35:20 ReplyNotPointerToStruct reply type not a pointer
to a struct: *int
2009/12/05 20:35:20 ReplyNotPublic reply type not public: *rpc.local
2009/12/05 20:35:20 rpc Register: type Bad has no public methods of
suitable type
FAIL

So I added:

println("reply: ", reply.C);

at line 92 of src/pkg/rpc/server_test.go just before:

if reply.C != args.A+args.B {
t.Errorf("Add: expected %d got %d", reply.C, args.A+args.B)
}

then the line in the test run:

Add: expected 4121272 got 15

goes away. I assumed the code generation changed and print caused
reply.C to be set correctly. The rest of the tests started to throw
exceptions.

It looks like the compiler is not generating correct code. I reverted
back to version 4329 and my build succeeds.

But wait, there's more. I then updated to 4330 and the build
succeeded, and then 4331, and 4332, and even tip succeeds. I figured
I'd go through the versions to see which change caused the issue. Now
my build works fine. Something really weird is happening...


geoff

[Risto Saarelma]

On Dec 5, 10:09 pm, Adam Langley <a...@golang.org> wrote:
> Does anyone else see this?

I'm getting this:

> $ cd src; ./all.bash
> ... many lines skipped ...
> make[2]: Entering directory `/home/rsaarelm/local/go/src/pkg/crypto/x509'
> 8g -o _gotest_.8 x509.go x509_test.go
> rm -f _test/crypto/x509.a
> gopack grc _test/crypto/x509.a _gotest_.8
> make[2]: Leaving directory `/home/rsaarelm/local/go/src/pkg/crypto/x509'
> --- FAIL: x509.TestCertificateParse
> cert not valid for host

> FAIL
> make[1]: *** [test] Error 1

> make[1]: Leaving directory `/home/rsaarelm/local/go/src/pkg/crypto/x509'

> make: *** [crypto/x509.test] Error 2
>

> $ hg identify
> bdfc3faa253a tip

> $ echo $GOOS $GOARCH
> linux 386

[baldmountain]

On Dec 5, 9:30 pm, baldmountain <baldmount...@gmail.com> wrote:
> Something really weird is happening...
>

I think this is caused by the unsafe pointer issue described by Ian
Lance Taylor in 386 instability on golang-dev.


geoff

[Russ Cox]

> I think this is caused by the unsafe pointer issue described by Ian
> Lance Taylor in 386 instability on golang-dev.

Indeed. Should be fixed now.

[Hugo Vincent]

Correct, fixed as of:
$ hg identify
53cc1c3cf1b6+ tip

Many thanks,
Hugo