bug#22354: Test failure when running distcheck from out-of-tree build

[taylanbayirli@gmail.com (Taylan Ulrich Bayırlı/Kammer)]

ludo@gnu.org (Ludovic Courtès) writes:

> taylanbayirli@gmail.com (Taylan Ulrich "Bayırlı/Kammer") skribis:
>
>> Yes, it still goes through the tests and fails at guix-environment.sh
>> with the same error.  I repeated all steps from a new clone of master.
>>
>> taylan@T420:~/src/guix/build$ grep length guix-0.9.1/_build/sub/config.log
>> configure:6946: checking the length of the installed socket file name
>> configure:6969: checking the length of the socket file name used in tests
>> configure:6980: checking the length of a typical hash bang line
>> configure:6992: checking the length of a hash bang line used in tests
>> ac_cv_guix_hash_bang_length=73
>> ac_cv_guix_socket_file_name_length=74
>> ac_cv_guix_test_hash_bang_length=127
>> ac_cv_guix_test_socket_file_name_length=90
>
> Arff, I see.  I’ve change LINUX_HASH_BANG_LIMIT in guix.m4 to 127
> instead of 128 to account for the trailing zero.
>
> This time it should detect the problem early.

Indeed, it fails with a meaningful error message at the beginning of the
check phase now.

> The “bad interpreter” message comes from Bash, in execute_cmd.c.
>
> However, that file does this:
>
>   char sample[80];
>
> […]
>
> #if defined (HAVE_HASH_BANG_EXEC)
> 	  READ_SAMPLE_BUF (command, sample, sample_len);
> 	  sample[sample_len - 1] = '\0';
> 	  if (sample_len > 2 && sample[0] == '#' && sample[1] == '!')
> 	    {
> 	      char *interp;
> 	      int ilen;
>
> 	      interp = getinterp (sample, sample_len, (int *)NULL);
>
> where READ_SAMPLE_BUF is:
>
> #define READ_SAMPLE_BUF(file, buf, len) \
>   do \
>     { \
>       fd = open(file, O_RDONLY); \
>       if (fd >= 0) \
> 	{ \
> 	  len = read (fd, buf, 80); \
> 	  close (fd); \
> 	} \
>       else \
> 	len = -1; \
>     } \
>   while (0)
>
> Now we know.  :-)

Yay for hard-coded limits! :P

> Ludo’.

Taylan