Return size of PDF page document

[hrdl-github]

This is a rough approximation to be used by the caching mechanism. I assume one byte per character and a small constant size for coordinates. I don't know how much overhead the table and individual structures require.

---

This change is 

[Frenzie]

lgtm

[Frenzie]

Hm, I guess that change causes this test to fail?

koreader-base/spec/unit/koptcontext_spec.lua

Line 203 in e2778f3

assert.are.same(kc_table.language, new_kc_table.language)

[hrdl-github]

I'm not sure why. I don't have a working test setup, but executing the test manually in luajit by pasting the relevant variables produces the desired result. Do you have an idea what is happening?

[benoit-pierre]

That test is flaky (because of possibly uninitialized memory).

[NiLuJe]

That test is flaky (because of possibly uninitialized memory).

Did you pinpoint exactly where the issue lies?

[Frenzie]

I told the tests to run again and they seem to have passed, so that would indeed seem to be slightly flaky. I'm not sure I've ever noticed it flaking before, or perhaps I just tune it out when I know it can't possibly be related which may not necessarily be the case here although it's unlikely.

[benoit-pierre]

That test is flaky (because of possibly uninitialized memory).

Did you pinpoint exactly where the issue lies?

Not yet.

[NiLuJe]

I told the tests to run again and they seem to have passed, so that would indeed seem to be slightly flaky. I'm not sure I've ever noticed it flaking before

It's been borked for years ;).

It used to be way way worse because at least one of the copies used to be more obviously borked (it wasn't null terminated), but it's still random today (and I think that's one of the rare "random" test failure that I could ever actually reproduce locally; but it's been a while since I tried [the fact that it requires setting up tesseract to run doesnt help ;p]).

So, yeah, something's fishy with that one, for sure.

[benoit-pierre]

See #1613.