[PATCH] posix: execvpe: skip $PATH components that are too long

[Pádraig Brady]

On 23/12/2025 18:22, Adhemerval Zanella Netto wrote:
> 
> I think it makes sense to handle ENAMETOOLONG as ENOENT, ESTALE, ENOTDIR,
> ENODEV, and ETIMEDOUT.  It is consistency with the systems (BSDs) and how
> other executors already handle it (shells).

...

> I think we also need to check if the output of the command execution
> does match what we expect ("foo\n"). Something like:
> 
> --
> 
> #include <stdlib.h>
> #include <libgen.h>
> #include <limits.h>
> 
> #include <support/capture_subprocess.h>
> #include <support/check.h>
> #include <support/support.h>
> #include <support/temp_file.h>
> #include <support/test-driver.h>
> #include <support/xunistd.h>
> 
> static char *fname;
> 
> static void
> do_prepare (int c, char** argv)
> {
>    int fd;
>    TEST_VERIFY_EXIT ((fd = create_temp_file ("testscript", &fname)) != -1);
> 
>    dprintf (fd, "echo foo\n");
>    TEST_VERIFY_EXIT (fchmod (fd, 0700) == 0);
>    xclose (fd);
> }
> #define PREPARE do_prepare
> 
> static void
> run_execvpe (void *closure)
> {
>    char *argv[] = { fname, NULL };
>    execvpe (basename (fname), argv, NULL);
> 
>    support_record_failure ();
> }
> 
> static int
> do_test (void)
> {
>    /* Linux 6.17 at least will give ENAMETOOLONG once any
>       path component is > NAME_MAX, so ensure we skip such paths
>       rather than failing.  [BZ #33626] */
>    char *path = xasprintf ("%*s:%s", NAME_MAX+1, "", test_dir);
>    TEST_VERIFY_EXIT (setenv ("PATH", path, 1) == 0);
>    free (path);
> 
>    struct support_capture_subprocess result
>      = support_capture_subprocess (run_execvpe, NULL);
>    support_capture_subprocess_check (&result, "execvpe", 0, sc_allow_stdout);
> 
>    TEST_COMPARE_STRING (result.out.buffer, "foo\n");
> 
>    support_capture_subprocess_free (&result);
> 
>    return 0;
> }
> 
> #include <support/test-driver.c>

Cool, I've updated the test with that in the attached.
I also fixed up the test file description.
The test passes, and fails with the code change reverted.

thanks!
Padraig
-------------- next part --------------
A non-text attachment was scrubbed...
Name: v3-glibc-execvpe-enametoolong.patch
Type: text/x-patch
Size: 4131 bytes
Desc: not available
URL: <https://sourceware.org/pipermail/libc-alpha/attachments/20251223/e5b2e1ab/attachment.bin>