Add CI

[eldruin]

I noticed one failing test for snprintf.
As the original author, is that something you could look into @thejpster?

Fixes #9

[jonathanpallant]

I think the test matrix is wrong. You can only pass the lp32 feature on platforms where Long and Pointers are 32-bit, which is not true on x86_64-unknown-linux-gnu. The difference in length is cause because -100i32 is passed to sprintf("%ld") and -100i32 isn't a long, and the wrong number of bytes is read giving a bad (much longer) string output.

[eldruin]

I see. Do you know any architecture available in cross that I can use to run the tests for LP32? What about the no-features case?

[jonathanpallant]

i686-something would be LP32 targets. You can target that on an x86-64 machine without using cross, I believe.

[jonathanpallant]

(I mean, you can /target/ anything but of course the point of the tests is to execute them, and an x86-64 machine will execute i686 code just fine).

[jonathanpallant]

I can't remember what happens with no features, and I'm afraid I'm not in a position to check right now. I seem to recall the code only checks LP64 and not(LP64) and only when looking at %ld for snprintf, but I could be wrong.

[thejpster]

Can you rebase this once the other PR is in? Maybe our issues will go away.

[eldruin]

They are indeed! This can now be reviewed and merged.

[thejpster]

Thanks!