| Summary: | sandbox mkdir wrapper differs in behavior than kernel | ||
|---|---|---|---|
| Product: | Portage Development | Reporter: | SpanKY <vapier> |
| Component: | Sandbox | Assignee: | Sandbox Maintainers <sandbox> |
| Status: | RESOLVED WORKSFORME | ||
| Severity: | normal | ||
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
it's been too long since i filed this bug, so i dont recall the behind-the-scene details. the proposed test case with `install` works fine with latest sandbox, so let's call it a day. |
the `make check` of coreutils fails because it tries to do: rm -rf d2 install -d d2/.. normal behavior would be: mkdir("d2/..", 0755) = -1 ENOENT (No such file or directory) mkdir("d2", 0755) = 0 mkdir("d2/..", 0755) = -1 EEXIST (File exists) stat("d2/..", {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0 <success!> the wrapper-funcs/mkdir.c does an lstat() on the canonicalized path rather than the original ... since canonicalize() doesnt actually do any filesystem calls, it has no idea if the specified path can even be resolved, it just assumes that it can be if i change the lstat() to operate on the original pathname rather than the canonicalized one, the `install -d d2/..` works just fine ...