Configure tests for 5.005 or later, but its getverlist uses a pack feature from 5.9.4

[gsteemso]

Description
Configure briefly stumbles over itself when running the getverlist script it constructs, because my ancient (version 5.8.6 or 5.8.8, depending on OS release) system Perls do not understand the ‘>’ endianness modifier in a pack template. I believe this causes no significant problems, but if it's completely ignoreable why would Configure bother running that code in the first place?

Steps to Reproduce
Run Configure alongside a pre-5.9-series Perl. I base this statement on the perlfunc documentation from version 5.12.5, which states that is when endianness modifiers were added to pack.

Expected behavior
Configure does not complain about invalid characters in a pack template.

Configuration: system Perl is 5.8.6 under Mac OS 10.4.11 and 5.8.8 under Mac OS 10.5.8. I could go into more detail but since I have to copy all this by hand into my iPhone browser and the issue is purely version dependent, I'm disinclined to go to all that effort unless someone has a really compelling reason.

[Tux]

This is a non-issue, as Configure is bound to the perl core distribution and not available as update or upgrade from CPAN

As current Configure is not supposed to work with any perl version older than the version it is shipped with, I think this can be closed as "wontfix"

[haarg]

This bit of Configure checks for an existing installed perl greater than version 5.005. It seems like that version check should be increased to the version that is actually needed.

[gsteemso]

“As current Configure is not supposed to work with any perl version older than the version it is shipped with”

...What? Configure is used for BUILDING the version of Perl it is shipped with, which by definition is not available yet.

[Tux]

mea culpa @gsteemso . I misinterpreted the situation. The code at hand in Configure is by me even.

": Determine list of previous versions to include in @inc"

would have helped in the problem description.

I agree with @haarg - 5.005 in @inc does not make sense at all

[Tux]

@@ -7685,6 +7685,9 @@ EOPL
 # of three digit per group separate by '.'s or '_'s. Names are extended
 # with ".0.0" to ensure at least three elements for the pack.
 #                                   -- H.Merijn Brand (m)'06 23-10-2006
+# pack "s>" was introduced in perl-5.10, so I had to rewrite this to an
+# sprintf approach to support bases systems with 5.6.x and 5.8.x
+#                                   -- H.Merijn Brand (m)'25 16-05-2025

 my @inc_version_list;
 my @candidates;
@@ -7692,7 +7695,7 @@ my @candidates;
 if (-d $stem) {
     chdir($stem);
     ;@candidates = map {
-       [ $_, pack "s>s>s>", split m/[._]/, "$_.0.0" ] } glob("5.*");
+       [ $_, sprintf "%04d%04d%04d", (split m/[._]/, "$_.0.0")[0,1,2] ] } glob("5.*");
     ;@candidates = sort { $a->[1] cmp $b->[1]} @candidates;
 }
 else {
@@ -7700,7 +7703,7 @@ else {
 }

 my ($pversion, $aversion, $vsn5005) = map {
-    pack "s>s>s>", split m/[._]/, "$_.0.0" } $version, $api_versionstring, "5.005";
+    sprintf "%04d%04d%04d", (split m/[._]/, "$_.0.0")[0,1,2] } $version, $api_versionstring, "5.005";
 foreach my $d (@candidates) {
     if ($d->[1] lt $pversion) {
        if ($d->[1] ge $aversion) {

Would that work for you?

[gsteemso]

Looks OK to me, but as I am sick in bed today, I am not in a position to test it.

[Tux]

@haarg Shall I merge that change? Does it conflict with the code freeze?

[haarg]

The change looks good to me, and I think it's fine to merge. The version of perl it checks for should still be updated. The addition of use warnings means it isn't compatible with perl 5.005 anymore.