secp256k throwing after windows build (32 bit go on win64)

[androlo]

Can't run on win64. Builds, but panics when secp256 function is called. Using MinGW32 gmp library as per the instructions. I have generated a working go executable using the gmp lib (5.1.2), so that part works. Output when running the secp256 tests:

=== RUN Test_Secp256_00
--- PASS: Test_Secp256_00 (0.00 seconds)
=== RUN Test_Secp256_01
2014/12/23 15:14:22 --- FAIL: Test_Secp256_01 (0.00 seconds)
panic:  [recovered]
        panic:

goroutine 21 [running]:
runtime.panic(0x5085a0, 0x124be118)
        C:/Go/src/pkg/runtime/panic.c:279 +0xe9
testing.func┬À006()
        C:/Go/src/pkg/testing/testing.go:416 +0x12d
runtime.panic(0x5085a0, 0x124be118)
        C:/Go/src/pkg/runtime/panic.c:248 +0x176
log.Panic(0x0, 0x0, 0x0)
        C:/Go/src/pkg/log/log.go:307 +0xa1
github.com/obscuren/secp256k1-go.VerifyPubkeyValidity(0x124ac4b0, 0x41, 0x41, 0x1)
        C:/Users/Andreas/go/src/github.com/obscuren/secp256k1-go/secp256.go:188 +0x38
github.com/obscuren/secp256k1-go.Test_Secp256_01(0x124ae6c0)
        C:/Users/Andreas/go/src/github.com/obscuren/secp256k1-go/secp256_test.go:45 +0xa6
testing.tRunner(0x124ae6c0, 0x5f292c)
        C:/Go/src/pkg/testing/testing.go:422 +0x87
created by testing.RunTests
        C:/Go/src/pkg/testing/testing.go:504 +0x720

goroutine 16 [chan receive]:
testing.RunTests(0x57ac04, 0x5f2920, 0x9, 0x9, 0x422101)
        C:/Go/src/pkg/testing/testing.go:505 +0x75c
testing.Main(0x57ac04, 0x5f2920, 0x9, 0x9, 0x5f8dc0, 0x0, 0x0, 0x5f8dc0, 0x0, 0x0)
        C:/Go/src/pkg/testing/testing.go:435 +0x6e
main.main()
        github.com/obscuren/secp256k1-go/_test/_testmain.go:63 +0x86

goroutine 17 [syscall]:
runtime.goexit()
        C:/Go/src/pkg/runtime/proc.c:1445

goroutine 19 [finalizer wait]:
runtime.park(0x412d30, 0x5f6998, 0x5f5c89)
        C:/Go/src/pkg/runtime/proc.c:1369 +0x94
runtime.parkunlock(0x5f6998, 0x5f5c89)
        C:/Go/src/pkg/runtime/proc.c:1385 +0x3f
runfinq()
        C:/Go/src/pkg/runtime/mgc0.c:2644 +0xc5
runtime.goexit()
        C:/Go/src/pkg/runtime/proc.c:1445
exit status 2
FAIL    github.com/obscuren/secp256k1-go        0.942s

Some info:

Windows 8.1, 64 bit
go 1.3.3, 32 bit
gcc 4.8.1
gmp-5.1.2-1-mingw32

[androlo]

Gonna do some experimentation myself. If someone has a working win32 build (or 64 for that matter) I would not mind getting their spec, so that I can test it.

[androlo]

Hmm turns out the secp256 tests throws on linux as well (ubuntu). at least for me. same error. newest libgmp-dev installed (fyi the build instruction says the wrong name, 'gmp-dev'). go 1.3.1 amd64 (ubuntu).

[androlo]

nm this.. it's the obscuren fork that doesn't pass. haltingstate does. this does not help the issue.

[androlo]

aaahaha, linking dynamically are we? Qt is fun! Oh well, closing (don't need that stuff).

[obscuren]

The original version of secp go has compressed pub keys. Ethereum requires uncompressed.

Fwiw, we are looking in to a native go secp256k1 implementation.

[androlo]

Yes, saw the diffs. Thank you for explaining.

[androlo]

Hey did you see that random generator fix committed @ haltingstate repo yesterday? Think that affects windows eth in any noticable way? Not a crypto guy myself so got a bit curious.

[obscuren]

@nicksavers have you seen this before?

@androlo No, I don't think so.

[androlo]

haltingstate was nice enough to drop by in our repo and set some things straight. he went over a lot of stuff. according to him, the rng issues comes from the fact that some platforms does not have a good source of entropy, not sure how it relates to the fix. He doesn't seem to like the uncompressed keys eth uses, though. If you want to see his notes they can be found here (along with some snarky layman commentary by yours truly): https://github.com/eris-ltd/thelonious/issues/67

[obscuren]

I've updated the sec256k1 repo which addresses the failing tests obscuren/secp256k1-go@2169cf8