Double bond stereo handling regression in version 1.07-beta.3

[IstvanOri]

Double bond stereo handling issues discoverd when v1.07-beta.3 was integrated. In Java environment, using the InChI native library via JNI, some double bond stereo related tests failed after upgrading to 1.07-beta3 from 1.06.

Only tested on windows yet, but it seems that not the Java integration is the issue, because can be reproduced (kind of) with the InChI command line tool (exe). Even though the command below runs succsessfully, the result seems incorrect.

How to reproduce

Run the following command (tested on windows):

inchi-1.exe in.txt /InChI2InChI

Where the content of in.txt file is InChI=1S/C4H8/c1-3-4-2/h3-4H,1-2H3/b4-3+

Current behaviour

The result is InChI=1S/C4H8/c1-3-4-2/h3-4H,1-2H3/b0-0!, which does not seem valid.

Expected behaviour

The result should be the original input.

[JanCBrammer]

@IstvanOri, thanks for the report! We'll look into this.

[djb-rwth]

I can confirm that the issue with /b-layer has now been fixed and will be added to Beta4 version very soon. Thanks for raising this issue.

[djb-rwth]

Hi Istvan,
v.1.07-Beta4 has been uploaded, so please check if the issue has been resolved.
Thank you.

[JanCBrammer]

@djb-rwth, could you link the closing commit?

[flange-ipb]

Closing commit is 4b84561.

The fix is in INCHI-1-SRC/INCHI_BASE/src/ichiread.c around line 7318.
The regression was introduced here before the 1.07-Beta1 release.

[IstvanOri]

Thank you for the fix. I can confirm it works.
It seems (by our automated tests) that in our system the InChI library could be upgraded to 1.07 in its current state.