On 03/12/20 22:57 -0600, Jacob Bachmeyer wrote:
Jonathan Wakely wrote:
On 03/12/20 13:33 +0000, Jonathan Wakely wrote:
The manual says "If DEJAGNU is set, but the file cannot be located, an
error will be raised and runtest will abort." That makes sense, if I
name a file that doesn't exist, I'm doing something wrong.
Despite the manual saying it's an error, and the comments in the code
saying it's an error, actually that's just a warning too.
# If $DEJAGNU isn't set either then there isn't any global
config file.
# Warn the user as there really should be one.
Sooooo, not really optional then?
Thus the warning. This code goes back to earliest snapshot in Git, so
I am unsure why it is handled this way.
if { ! [info exists env(DEJAGNU)] } {
send_error "WARNING: Couldn't find the global config file.\n"
}
if { [load_file -- $env(DEJAGNU)] == 0 } {
# It may seem odd to only issue a warning if there isn't a global
# config file, but issue an error if $DEJAGNU is erroneously defined.
# Since $DEJAGNU is set there is *supposed* to be a global
config file,
# so the current behaviour seems reasonable.
send_error "WARNING: global config file $env(DEJAGNU) not found.\n"
}
That looks pretty similar to the case where $DEJAGNU isn't set at all.
What's going on here?
The difference is that if $DEJAGNU *is* set but the file does not
exist, the user evidently intended to load a global configuration
file, while if $DEJAGNU is not set at all, the user presumably does
not care. The current master (as of commit
660af948c8fce93ad2781ab8e0527a5b8e216f14) declares an "ERROR" and
exits instead of emitting a pointless warning and continuing in this
case.
Ah good, so now it actually matches the comment in the code and the
documentation.