Assert on NULL IO functions

[wangdongustc]

Assert that the IO functions (especially the sync function when using non-caching storage) are provided by the user.

[geky-bot]

Tests passed ✓, Code: 17128 B, Stack: 1440 B, Structs: 812 B

	Code	Stack	Structs		Coverage
Default	17128 B	1440 B	812 B	Lines	2396/2568 lines
Readonly	6250 B	448 B	812 B	Branches	1280/1610 branches
Threadsafe	17976 B	1440 B	820 B		Benchmarks
Multiversion	17200 B	1440 B	816 B	Readed	29369693876 B
Migrate	18816 B	1736 B	816 B	Proged	1482874766 B
Error-asserts	17872 B	1432 B	812 B	Erased	1568888832 B

[geky]

Hi @wangdongustc, thanks for creating a PR.

I think this would be more useful in lfs_init, which is where we have other config-related asserts. lfs_init is always called before any filesystem operations.

Are you able to also include asserts on read/prog/erase being non-null? These should all probably follow the same rules.

But only if LFS_READONLY is not defined:

LFS_ASSERT(lfs->cfg->read != NULL);
#ifndef LFS_READONLY
LFS_ASSERT(lfs->cfg->prog != NULL);
LFS_ASSERT(lfs->cfg->erase != NULL);
LFS_ASSERT(lfs->cfg->sync != NULL);
#endif

This gets quite a bit more nuanced than it first appears...

---

And clever solution for the prettyasserts.py issue, thanks for that.

I had a fix for pointer comparisons on an unrelated branch (b5e264b), but I like your solution more.

(We will probably want to end up with both eventually to allow pointer comparisons without a literal NULL)

[wangdongustc]

@geky Thanks for the feedback!

Are you able to also include asserts on read/prog/erase being non-null? These should all probably follow the same rules.

Done!

(Actually I did this earlier yesterday. But then I thought, why bother asserting the "read" function? Nobody would expect that lfs will work without a "read" function. So I changed back to only assert the "sync" function later. 🤣

[geky-bot]

Tests passed ✓, Code: 17128 B, Stack: 1440 B, Structs: 812 B

	Code	Stack	Structs		Coverage
Default	17128 B	1440 B	812 B	Lines	2399/2571 lines
Readonly	6250 B	448 B	812 B	Branches	1283/1616 branches
Threadsafe	17976 B	1440 B	820 B		Benchmarks
Multiversion	17200 B	1440 B	816 B	Readed	29369693876 B
Migrate	18816 B	1736 B	816 B	Proged	1482874766 B
Error-asserts	17892 B	1432 B	812 B	Erased	1568888832 B

[geky]

But then I thought, why bother asserting the "read" function? Nobody would expect that lfs will work without a "read" function.

Now I'm wondering what it would take to support write-only memory 😬

Thanks for this! Merging!

[bmcdonnell-fb]

write-only memory 😬

WOM. 😆

[geky]

https://web.archive.org/web/20160313132244/https://dl.dropboxusercontent.com/u/85027770/25120SpecTwoSides%20NC.pdf 😄

Apparently someone wrote a kernel driver for it: https://github.com/spacerace/write-only-memory