Support server paths #1136
Support server paths #1136
Conversation
| @@ -67,6 +67,14 @@ int path_getrelative(lua_State* L) | |||
| return 1; | |||
| } | |||
|
|
|||
| /* Relative paths within a server can't climb outside the server root. | |||
| * If the paths don't share server name, return the absolute path. */ | |||
| if (src[0] == '/' && src[1] == '/' && last == 1) { | |||
There was a problem hiding this comment.
I feel that this might overrun the buffer if one were to attempt to get a relative path for "". To avoid that, this could probably be:
if (src[0] != '\0' && src[0] == '/' && src[1] == '/' && last == 1) {There was a problem hiding this comment.
But isn't src[0] != '\0' && src[0] == '/' redundant? if src[0] == '/' then it's guaranteed to not be 0
There was a problem hiding this comment.
Additionally, empty strings are covered in the previous check.
There was a problem hiding this comment.
Haha, not really sure what I was thinking when I made that comment. Sorry about that!
| /* add to the result, filtering out duplicate slashes */ | ||
| if (ch != '/' || last != '/') { | ||
| /* add to the result, filtering out duplicate slashes, except when they are leading slashes */ | ||
| if (str == &source[1] || (ch != '/' || last != '/')) { |
There was a problem hiding this comment.
I think this is the line I was meant to comment on. Is source guaranteed to not be an empty string here?
There was a problem hiding this comment.
Without testing, I believe so. if the string is empty, then str will be equal to endPtr and the loop condition while (str != endPtr) will fail.
There was a problem hiding this comment.
After some testing I found out that it never attempt to normalize an empty string, thanks to
premake-core/src/host/path_normalize.c
Line 108 in a713ecc
For context, see #1129
TLDR: Server paths (i.e. "//server/files") were treated like any non-normalized path with duplicate slashes.