Add a 'target' field for symlinkattributes

[hishamhm]

It returns the resolved path of the symlink.

Since this adds functionality, this should be merged in the 1.7 branch. PRing for review.

[coveralls]

Coverage increased (+1.04%) to 80.102% when pulling 2e068eb on symlink_target into 81e5b16 on 1.7.

[hishamhm]

(I don't know why these other commits 6e0dc7b and 7dae11c ended up in this PR too, but I guess they should be in 1.7 too)

[coveralls]

Coverage increased (+1.09%) to 80.151% when pulling a1015fe on symlink_target into 81e5b16 on 1.7.

[daurnimator]

Why not implement on windows?

[hishamhm]

Be my guest! :)

[daurnimator]

So that implies it was just a lack of time rather than a technical obstacle?

[hishamhm]

...and interest — I usually don't do Windows coding for free :)

[hishamhm]

For cross-referencing: I noticed that this closes #43.

[mpeterv]

(I don't know why these other commits 6e0dc7b and 7dae11c ended up in this PR too, but I guess they should be in 1.7 too)

Probably because I haven't merged master into 1.7 yet and this PR's branch was created from master.

[mpeterv]

Not sure what is going on with LuaJIT failures, see mpeterv/hererocks#27.

[mpeterv]

@hishamhm I don't like that to getting size for readlink requires indexing a table that was just built. Maybe could return stat struct (using pointer argument) to make post-processing easier?

[mpeterv]

Alternatively, it should be possible to use existing mechanism for declaring functions pushing fields. But then there has to be a way to use different field lists for stat and lstat, and also these pushing functions should be able to signal failure.

[n1tehawk]

I'd like to suggest a different approach for determining the buffer size:

Start with a reasonable initial value (say 256 or 512), and check the readlink() return value against it. If that suggests that the buffer may have been filled completely, realloc() it to a bigger size and retry. See n1tehawk@3497406#diff-f6b7bcefa5f97fba6fa9e82b43357724R865 ...

If you wish, I'll create a pull request for this.
Regards, NiteHawk

[EDIT] P.S.: I realized that size - 1 isn't needed, and fixed my branch accordingly: n1tehawk@4185914

[hishamhm]

@n1tehawk yes, this PR would be welcome!

[n1tehawk]

See #78.

Compared to my previous suggestion, I have added proper readlink() error handling. Also: The size restriction was too conservative/narrow without need - since we'll only write target[tsize] if (tsize < size), it's safe to use size for the readlink() call directly (instead of size - 1).

Regards, NiteHawk

[n1tehawk]

This PR was superseded by merging #78, and should be closed.

Regards, NiteHawk