Smalltalk › Gnu

[PATCH 3/3] cint: Fail the VMpr_CFuncDescriptor_call primitive with wrong args

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

6 messages Options

Holger Freyther

[PATCH 3/3] cint: Fail the VMpr_CFuncDescriptor_call primitive with wrong args

Instead of crashing the VM we will fail the primitive if a type
can not be converted from Smalltalk to C. The code attempts to
free all data allocated for this primitive.

2011-02-04 Holger Hans Peter Freyther <[hidden email]>

* libgst/cint.c: Propagate type conversion failures.

2011-02-04 Holger Hans Peter Freyther <[hidden email]>

* UnitTest.st: Verify that the primitive are failing.

Signed-off-by: Holger Hans Peter Freyther <[hidden email]>
---
ChangeLog | 4 +++
libgst/cint.c | 51 +++++++++++++++++++++++++++++------------
packages/sockets/ChangeLog | 4 +++
packages/sockets/UnitTest.st | 15 ++++++++++++
4 files changed, 59 insertions(+), 15 deletions(-)

_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

0003-cint-Fail-the-VMpr_CFuncDescriptor_call-primitive-wi.patch (5K) Download Attachment

Holger Freyther

Re: [PATCH 3/3] cint: Fail the VMpr_CFuncDescriptor_call primitive with wrong args

On 02/04/2011 10:28 PM, Holger Hans Peter Freyther wrote:
> +cleanup:
> + /* Attempt to cleanup all pushing of arguments. We have tried to
> + write at si and now need to cleanup up until this point. */
> + for (i = 0, arg = local_arg_vec; i < si; i++, arg++)

okay the < si is wrong...i should probably use c_func_cur->arg_idx or such. I
will create another testcase that can be used to verify we are not leaking
memory here.

_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

Holger Freyther

Re: [PATCH 3/3] cint: Fail the VMpr_CFuncDescriptor_call primitive with wrong args

In reply to this post by Holger Freyther

On 02/04/2011 10:28 PM, Holger Hans Peter Freyther wrote:

> + testDoNotCrashWithWrongTypes [
> + "The objective is to see if wrong types for a cCallout will
> + make the VM crash or not. It should also check if these calls
> + raise the appropriate exception."
> + | socket impl |
> +
> + socket := DatagramSocket new.
> + impl := socket implementation.
> +
> + self should: [impl accept: -1 peer: nil addrLen: 0] raise: SystemExceptions.PrimitiveFailed.
> + self should: [impl getPeerName: -1 addr: nil addrLen: 0] raise: SystemExceptions.PrimitiveFailed.
> + self should: [impl getSockName: -1 addr: nil addrLen: 0] raise: SystemExceptions.PrimitiveFailed.
> + self should: [impl receive: -1 buffer: nil size: 0 flags: 0 from: nil size: 0] raise: SystemExceptions.PrimitiveFailed.
> + ]
> ]

this fails from within the testsuite as it is generating stderr output. Is
this something we can handle in autotest?

_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

Paolo Bonzini-2

Re: [PATCH 3/3] cint: Fail the VMpr_CFuncDescriptor_call primitive with wrong args

On 02/04/2011 10:52 PM, Holger Hans Peter Freyther wrote:

> On 02/04/2011 10:28 PM, Holger Hans Peter Freyther wrote:
>> + testDoNotCrashWithWrongTypes [
>> + "The objective is to see if wrong types for a cCallout will
>> + make the VM crash or not. It should also check if these calls
>> + raise the appropriate exception."
>> + | socket impl |
>> +
>> + socket := DatagramSocket new.
>> + impl := socket implementation.
>> +
>> + self should: [impl accept: -1 peer: nil addrLen: 0] raise: SystemExceptions.PrimitiveFailed.
>> + self should: [impl getPeerName: -1 addr: nil addrLen: 0] raise: SystemExceptions.PrimitiveFailed.
>> + self should: [impl getSockName: -1 addr: nil addrLen: 0] raise: SystemExceptions.PrimitiveFailed.
>> + self should: [impl receive: -1 buffer: nil size: 0 flags: 0 from: nil size: 0] raise: SystemExceptions.PrimitiveFailed.
>> + ]
>> ]
>
> this fails from within the testsuite as it is generating stderr output. Is
> this something we can handle in autotest?

I'll make it XFAIL for now. For 3.3 we can change the PrimitiveFailed
to invoke a Smalltalk implementation of the type checks and raise a
proper exception.

Paolo

_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

Paolo Bonzini-2

Re: [PATCH 3/3] cint: Fail the VMpr_CFuncDescriptor_call primitive with wrong args

In reply to this post by Holger Freyther

On 02/04/2011 10:40 PM, Holger Hans Peter Freyther wrote:
> okay the< si is wrong...i should probably use c_func_cur->arg_idx or such. I
> will create another testcase that can be used to verify we are not leaking
> memory here.

Please send an incremental 4/3 patch and I'll squash it into this one.
Thanks!

Paolo

_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

Holger Freyther

Re: [PATCH 3/3] cint: Fail the VMpr_CFuncDescriptor_call primitive with wrong args

In reply to this post by Paolo Bonzini-2

On 02/05/2011 04:29 PM, Paolo Bonzini wrote:

> I'll make it XFAIL for now. For 3.3 we can change the PrimitiveFailed to
> invoke a Smalltalk implementation of the type checks and raise a proper
> exception.

It fails because we don't ignore the stderr output... e.g. this will make the
test case pass. I will put that into another patch... so you can put this
before the cint patch.

diff --git a/tests/local.at b/tests/local.at
index 49f9111..7dedf2c 100755
--- a/tests/local.at
+++ b/tests/local.at
@@ -64,7 +64,7 @@ m4_define([AT_OPTIONAL_PACKAGE_TEST], [
case $ret in
2) exit 77 ;;
0|1) exit $ret ;;
- esac], [], [], [ignore])
+ esac], [], [], [ignore], [ignore])
AT_CLEANUP
])

_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk