Here's the fix for DBI 1.16 to work with perl 5.004:
***************
*** 2516,2522 ****
SvREFCNT_dec(hv); /* since newRV incremented it */
SvREFCNT_dec(ka_rv); /* since we created it */
}
! else { RETVAL = &sv_undef; }
OUTPUT:
RETVAL
--- 2515,2526 ----
SvREFCNT_dec(hv); /* since newRV incremented it */
SvREFCNT_dec(ka_rv); /* since we created it */
}
! else {
! RETVAL = &sv_undef;
! #if (PERL_VERSION < 4) || ((PERL_VERSION == 4) && (PERL_SUBVERSION <= 4))
! RETVAL = newSV(0); /* mutable undef for 5.004_04 */
! #endif
! }
OUTPUT:
RETVAL
Tim.
On Thu, May 31, 2001 at 01:10:09PM +0200, Merijn Broeren wrote:
> Hi,
>
> Reading the DBI 1.16 README it states :
>
> Build, test and install Perl 5 (preferably 5.004_04 or later).
>
> But when I try to build and test DBI 1.16 with 5.00404 it breaks
> during testing like this:
>
> Modification of a read-only value attempted at blib/lib/DBI.pm line
> 1130.
>
> This happens consistently on Solaris 2.5.1 & 2.7, Irix 6.5 and
> Linux 2.2 & 2.4. All tests are successful with perl 5.00503 and
> perl 5.6.1 on all these platforms.
>
> The actual test output looks like this :
>
> t/basics............ok
> t/dbidrv............ok
> t/examp.............Modification of a read-only value attempted at
> blib/lib/DBI.pm line 1130.
> dubious
> Test returned status 2 (wstat 512, 0x200)
> DIED. FAILED tests 65-127
> Failed 63/127 tests, 50.39% okay
> t/meta..............ok
> t/proxy.............skipping test on this platform
> t/shell.............skipping test on this platform
> t/subclass..........ok
> Failed Test Status Wstat Total Fail Failed List of failed
> -------------------------------------------------------------------------------
> t/examp.t 2 512 127 63 49.61% 65-127
> Failed 1/7 test scripts, 85.71% okay. 63/185 subtests failed, 65.95%
> okay.
>
> If I edit t/examp.t to comment out the mortal failures I have
> to comment out several calls to fetchall_arrayref, selectall_hashref
> and fetchrow_hashref, but not all. The 'Modification of a read-only
> variable' error varies in place with the call origination.
>
> After this undiscerning commenting out in t/examp.t I see the following
> result :
>
> t/examp.............
> # failed test 67 at line 215 (row count 0)
> # failed test 68 at line 216 (0 vs 31)
> # failed test 78 at line 234
> # failed test 80 at line 244
> DBI handle cleared whilst still active.
> dbih_clearcom (h 0x21a0a4, com 0x230f08):
> FLAGS 0x1011: COMSET Warn Taint
> TYPE 3
> PARENT DBI::db=HASH(0x24fc8c)
> KIDS 0 (0 active)
> IMP_DATA ARRAY(0x25182c) in 'DBD::ExampleP::st'
> NUM_OF_FIELDS 3
> NUM_OF_PARAMS 1
> FAILED tests 67-68, 78, 80
> Failed 4/115 tests, 96.52% okay
>
> I tried to add a explicit ->disconnect to the database handles
> to dispose of the 'cleared whilst still active', to no avail.
>
> So, is the README wrong? Do we need at least 5.00405 or even 5.00503
> to work with 1.16? I don't mind that much, it is a good incentive to
> move our developers to 5.005 and 5.6, but it needs to be mentioned
> at least.
>
> Regards,
> --
> Merijn Broeren |
> Software Geek | Yield to temptation; it may not pass your way again.
> |
