Looking up words on this subtitle causes segmentation fault about memento

[Calvin-Xu]

I thought I found another string that could trigger this:

〝新型砕氷艦の
就航にともない―〞

but testing it one last time before submitting it, I suddenly could not reproduce the crash with either string. If this bug could not be reproduced at all on your end, I think I should close this issue.

[Calvin-Xu]

I now understand why I could not reproduce the issue last time. During the interval, I changed the system dictionary of my MeCab installation from UniDic back to IPAdic in my mecabrc. The crash only happens when UniDic is used.

Since Memento comes bundled with / depends on IPAdic by default ("with a system dictionary like ipadic on Linux and OS X"), I think this only affects users who already use their local MeCab with UniDic (or possibly other dictionaries).

On another note, I had thought about proposing switching the bundled dictionary to UniDic (and tried the changes in a fork), since it is the actively maintained dictionary. But it does come at the cost of larger bundle sizes, one data file that exceeds GitHub's 100MB limit (UniDic v2.1.2), and that no mecab-unidic is currently available on AUR.

[ripose-jp]

Not that it's ideal, but I'm only aware of Memento working with ipadic and NAIST-jdic. I believe way way back I tested out UniDic when choosing a dictionary to bundle with the Windows version and decided against UniDic because it was way too big to upload to GitHub without LFS and it caused crashes because it doesn't follow the exact same format as ipadic.

From my understanding, passing a -O argument to MeCab when initializing it might be able to fix it, but I couldn't find any documentation on this option, so I just didn't bother.

MeCab dictionaries shouldn't be user changable in 99% of cases, so this is a bit of a low priority bug.

[Calvin-Xu]

Thanks. I suppose it might be a good idea to close this issue for now. But perhaps we should note somewhere that Memento might crash with dictionaries other than ipadic and NAIST-jdic (possibly in the build section?) since currently the documentation doesn’t really advise against it.

[ripose-jp]

I've clarified the language in the README under the Dependencies section.