StdCallCallback fails with message of corrupted stack

[boscogh]

I found out use case there StdCallCallback fails with message of corrupted stack.
It only occure with method of specific signature.
Please review my question on stackoverflow, it contains isolated testcase to 100% reproducible issue

link to the question details with source code included:
http://stackoverflow.com/questions/21039368/jna-stdcall-esp-corruption

[twall]

Take a look at branch issue-300 to see if you can tweak it to reproduce the error you're seeing.

[twall]

9 arguments works, 11 arguments doesn't. Branch updated.

[twall]

libffi may be calculating an incorrect stack size (actual 60, expected 56). This number is inserted into a dynamically generated trampoline, defined in libffi/src/x86/ffi.c.

My expectation may be off.

[twall]

Updated binaries on issue-300 branch. See if it fixes your issue.

[boscogh]

just tested, works perfect, thank you very much!
would it be included in some next release?
I would like to post solution in the forum of provider of that 3rd party dll.
folks tried to run it through JNA unsuccesfully since 2008 year ;-)

[twall]

It’ll be included in the next release. It only affects w32-x86.

On Jan 13, 2014, at 2:57 AM, boscogh [email protected] wrote:

just tested, works perfect, thank you very much!
would it be included in some next release?
I would like to post solution in the forum of provider of that 3rd party dll.
folks tried to run it through JNA unsuccesfully since 2008 year ;-)

—
Reply to this email directly or view it on GitHub.

[twall]

BTW, this is a bug in the underlying libffi library, so a patch will be forwarded to that project as well.

On Jan 13, 2014, at 2:57 AM, boscogh [email protected] wrote:

just tested, works perfect, thank you very much!
would it be included in some next release?
I would like to post solution in the forum of provider of that 3rd party dll.
folks tried to run it through JNA unsuccesfully since 2008 year ;-)

—
Reply to this email directly or view it on GitHub.

[boscogh]

yes, I saw the sources of your patch and I already posted on the forum of
the 3rd party DLL (which was a part of Quik API, btw) that this was a bug
in libffi.
it is very cool to see such a fast patch, I am really glad to see such a
brilliant job!

2014/1/13 Timothy Wall [email protected]

BTW, this is a bug in the underlying libffi library, so a patch will be
forwarded to that project as well.

On Jan 13, 2014, at 2:57 AM, boscogh [email protected] wrote:

just tested, works perfect, thank you very much!
would it be included in some next release?
I would like to post solution in the forum of provider of that 3rd party
dll.
folks tried to run it through JNA unsuccesfully since 2008 year ;-)

Reply to this email directly or view it on GitHub.

Reply to this email directly or view it on GitHubhttps://github.com//issues/300#issuecomment-32165077
.

Ó Õ×ÁÖÅÎÉÅÍ,
ðÁ×ÅÌ íÁÌÀÔÉÎ