The Guile build I posted about yesterday did complete successfully after the 
three edits I described in that post.

However, the test suite invoked by a subsequent `make check` fails in a few 
places, and I do not know enough about Guile to tell whether any of these are 
indicative of deep problems or if they are all relatively trivial.  
Specifically:

- In `00-repl-server.test`, I got an intermittent error, “thread_suspend 
failed”.  It happened once, with no further information provided and no obvious 
cause.

- In `00-socket.test`, the `setsockopt` test fails outright, while the `bind 
abstract socket` test errors out with “no such file or directory” (which then 
causes the subsequent `listen abstract`, `connect abstract`, and `accept 
abstract` tests to not be able to resolve).  These all have 100% repeatability.

- A significant minority of test scripts reported unresolved tests, but no 
actual errors or failures.  `i18n.test` in particular produced 43 unresolved 
tests, well over half of which involved French.  Strangely, the summary of test 
results appears to treat them as successes, unless I am misinterpreting 
something.

- In `filesys.test`, `mkdtemp` fails outright with an “invalid template”, while 
`chmodat` evokes three separate errors:  “not a port”, “not a file port”, and 
“closed port” (all three errors speak of an “unbound variable #f”.)

- In `posix.test`, there are two errors (“utime: AT_SYMLINK_NOFOLLOW” has an 
out-of-range argument 4096; “pipe: O_CLOEXEC” has an unknown error 0), three 
unresolved tests, and an outright failure (“spawn: wc with stdin and stdout 
redirects”) so severe as to prompt a backtrace (which reveals that the “2” it 
reports expecting should in fact have been a Unicode HORIZONTAL ELLIPSIS).

- Running `r4rs.test` produced an interpreter warning about “load” being used 
in a declarative module.

- Between `srfi-98.test` and `srfi-105.test`, an interpreter warning appeared:  
“call to environ while multiple threads are running; further behavior 
unspecified”.

Any commentary shedding light on these errors/failures would be appreciated.

Gordon S.

Reply via email to