52 twolc test pairs with uneven strings ( about lang-sms HOT 2 OPEN

Comment 11136

Date: 2016-02-15 11:17:25 +0100
From: Sjur Nørstebø Moshagen <<sjur.n.moshagen>>

In an effort to move forward with bug #2061, I tried compiling and testing sms on different systems. SMS does compile on my system, with both Hfst and Xerox, but running make check gives a rather unwanted output:

• Using Hfst, the hfst-test-pair tool crashes with an unfriendly error message
• Using Xerox, there are a lot of useless errors because of uneven strings

I believe the Hfst error is caused by the same issue, just that the error message is not as nice as it could be.

There are 52 such uneven strings. Please correct them before doing more work on twolc, so that we have a baseline to compare things against when trying to work out the issues with bug #2061.

Also, I suggest you compile and test using Xerox in addition to (or instead of) Hfst for the time being. It is much faster (10 minutes including documentation vs more than 30 mins on my machine), and the twolc testing error messages are much more informative when it comes to uneven strings.

from lang-sms.

Comment 11143

Date: 2016-02-17 18:14:20 +0100
From: Sjur Nørstebø Moshagen <<sjur.n.moshagen>>

After Jack's recent changes sms is now down to 12 such test pairs with an uneven number of symbols :-)

I belive Miikka fixed the crashing bug in hfst-test-pairs, so that it now instead gives an error message similar to Xerox, but I have not tested that.

from lang-sms.