[2; .txt, .js] Outputs entirety of generated file despite there being only two changes; tries to allocate 16.3 TiB about difftastic

[Wilfred]

lojbo .ui

Thanks for the report.

I think the first issue is probably trying to diff lines ending \r\n against files ending with just \n and concluding that they're different.

The second issue is probably due to line diffing trying to highlight words within the same line.

[Wilfred]

I can repro the second issue, but not the first. Can you reproduce the first issue with local copies of the files? If so, could you attach them here?

[Wilfred]

OK, I can reproduce the first issue if I have two files that differ by line ending, but I'm not convinced that difftastic's behaviour is wrong here.

$ echo "one\ntwo" > a.txt                            
$ echo "one\nfoo\ntwo" > b.txt
$ unix2dos b.txt       

$ difft a.txt b.txt
b.txt --- Text
1 one                             1 one
.                                 2 foo
2 two                             3 two

$ diff a.txt b.txt 
1,2c1,3
< one
< two
---
> one
> foo
> two

Plain GNU diff also considers these files to be completely different. Do you have any special crlf settings in your Windows git setup?

[berrymot]

I can repro the second issue, but not the first. Can you reproduce the first issue with local copies of the files? If so, could you attach them here?

Unfortunately I'll be away from my computer until next week, sorry. I don't think it's the line endings though? Pretty sure allwords.txt has been CRLF all along. When I get back I'll try adding a test entry to the database and regenerate allwords.txt and see if this still happens.

As for

Do you have any special crlf settings in your Windows git setup?

I don't remember ever messing with that- the one thing I know I've done to etc/config is using difftastic rather than the default diff lol

ni'o lu «lojbo .ui» li'u zo'u .ue xu do jbopre

[berrymot]

The test entry has been made, it happens to mean 'line terminator' lol; I'll rerun the parsing stuff when I can

[berrymot]

Reran the script with the new word.

allwords.txt still printed the entire file:

data.txt still didn't:

Replaced every \n with \r\n in the script, reran.

This fixed allwords.txt, but data.txt's changes were too big.

Reverted the changes to those two, manually replaced each LF with CRLF, committed, reran.

wheeeeeee

[berrymot]

Updated to 0.56.1, turns out printing such a giant diff is REALLY bad for my terminal lol