On Wed, Feb 15, 2023 at 8:55 AM Jim Jones <jim.jo...@uni-muenster.de> wrote:
>
> On 13.02.23 13:15, Jim Jones wrote:
>
> diff -U3 /tmp/cirrus-ci-build/src/test/regress/expected/xml.out
> /tmp/cirrus-ci-build/build/testrun/regress/regress/results/xml.out
> --- /tmp/cirrus-ci-build/src/test/regress/expected/xml.out 2023-02-12
> 09:02:57.077569000 +0000
> +++ /tmp/cirrus-ci-build/build/testrun/regress/regress/results/xml.out
> 2023-02-12 09:05:45.148100000 +0000
> @@ -1695,10 +1695,7 @@
> -- XML format: empty string
> SELECT xmlformat('');
> ERROR: invalid XML document
> -DETAIL: line 1: switching encoding : no input
> -
> -^
> -line 1: Document is empty
> +DETAIL: line 1: Document is empty
>
> ^
> -- XML format: invalid string (whitespaces)
>
> I couldn't figure out why the error messages are different -- I'm wondering
> if the issue is the test environment itself. I just removed the troubling
> test case for now
>
> SELECT xmlformat('');
>
> v7 attached.
>
> Thanks for reviewing this patch!
>
Yesterday I looked at those cfbot configs and noticed all those
machines have different versions of libxml.
2.10.3
2.6.23
2.9.10
2.9.13
But I don't if version numbers have any impact on the different error
details or not.
~
The thing that puzzled me most is that in MY environment (CentOS7;
libxml 20901; PG --with-libxml build) I get this behaviour.
- Without your v6 patch 'make check' is all OK.
- With your v6 patch other XML tests (not only yours) of 'make check'
failed with different error messages.
- Similarly, if I keep the v6 patch but just change (in xmlformat) the
#ifdef USE_LIBXML to be #if 0, then only the new xmlformat tests fail,
but the other XML tests are working OK again.
Those results implied to me that this function code (in my environment
anyway) is somehow introducing a side effect causing the *other* XML
tests to fail.
But so far I was unable to identify the reason. Sorry, I don't know
this XML API well enough to help more.
------
Kind Regards,
Peter Smith.
Fujitsu Austalia.
