Capitalized Cyrillic word not working with lt-proc

[rmlockwood]

bilingual.txt

The attached bilingual dictionary has the line:
<e w="1"><p><l>наб1.1<s n="perspron" /></l><r>дам1.1<s n="pers" /></r></p></e>
I first compile it like this:
lt-comp lr bilingual.txt bilingual.bin

When I run lt-proc with this word uncapitalized it works fine:

echo "^наб1.1<perspron>$" |lt-proc -b -N1 -L1 bilingual.bin
"^наб1.1<perspron>/дам1.1<pers>$"

When I run lt-proc with the capitalized version, it fails to find the lowercase equivalent of the word in the bilingual dictionary:

>echo "^Наб1.1<perspron>$" |lt-proc -b -N1 -L1 bilingual.bin
"^Наб1.1<perspron>/@Наб1.1<perspron>$"

Here's some version info:

lt-proc -v
lt-proc version 3.8.1
lt-comp
lt-comp v3.8.1: build a letter transducer from a dictionary

I am using the Windows build of these tools.
An Apertium developer colleague said he could not reproduce this in Linux see: rmlockwood/FLExTrans#1111

[unhammer]

yeah, seems windows-specific like @mr-martian said:

$ lt-comp lr bilingual.txt bilingual.bin 
append@standard 16 15
main@standard 76 102
$ echo "^наб1.1<perspron>$" |lt-proc -b -N1 -L1 foo.bin 
^наб1.1<perspron>/дам1.1<pers>$
$ echo "^Наб1.1<perspron>$" |lt-proc -b -N1 -L1 foo.bin 
^Наб1.1<perspron>/Дам1.1<pers>$
$ echo "^НАБ1.1<perspron>$" |lt-proc -b -N1 -L1 foo.bin 
^НАБ1.1<perspron>/ДАМ1.1<pers>$

[unhammer]

this is a "real" windows build, not WSL? How did you build it?

[unhammer]

(also, my lt-proc says 3.8.2, not 3.8.1)

[mr-martian]

He's using the Windows builds from nightly, and in his current set up, updating is a bit complicated, so he's often a few patches behind main.

[rmlockwood]

Can someone look into this problem?

[rmlockwood]

I'm having the same issue with version 3.8.2 of lt-proc

[TinoDidriksen]

I can reproduce this with the binaries built via MXE. Binaries built with Visual Studio work, so looks like I have some debugging/scripting to do.

[rmlockwood]

I would really appreciate if you could debug this!

[rmlockwood]

Any progress?