Hi developers
My name is Josef.
I am porting, actually only building libxml2 2.6.27 on a HP Alpha OpenVMS System.
So far no problems. I have made a couple of mmk make files to describe my system,
the library and all the test suite files for a automatic make like rebuild on OpenVMS.
At the moment I am about to start testing using RUNTEST.c
On the PC I have build libxml2 using the MS Visual Studio V6.0 's IDE , not just nmake
All went fine. after setting up the full bunch of projects.
The problem starts when I run RUNTEST. on the PC as well as on the OpenVMS System.
The dumped files *.res are almost ever different from the files residing under ./result, to which the
*.res files are compared.
meanwhile I have found the following:
catching tghe ./test and ./result files from the SVN trunk gives me by far the best results when I run
regression tests on MSVC / IDE (nearly .-( faultfree)
moving the ./result and ./test files to my OpenVMS system using NetBeans and Distributed NetBeans
(on top of an FTP service) I found several formats at [.result] on the OpenVMS System
Again, taking the files from a mirror which delivers to i.e. Solaris is a nigth mar, getting it from the SVN trunk
is much better, far from perfect.
When I move the files from PC to OpenVMS (to a ODS5 Voume) some files come in with variable record length,
some come in with fixed record lengths. i.e. the ./result/dave1 shows after the last closing tag hundres of little blocks
[][][][][][][][][][][][][][][][][][][][][][][][]
When I delet all this blocks using NetBeans, then put them to the OpenVMS Server, and then do
$ convert/fdl=convert.fdl dav1. dav1.
the format is then changed from fix length records to a stream_LF format
after that I can run the regression test for ./result/dav1. and dav1.res perfectly still parsing from ./test/dav1
It is a nightamr
The problems I am partly faced with is as such and found in runtest.c when xmlSaveFile(temp, doc); gets called
xmlSaveFile from xmlsave.c dumps the doc to a real file and appends a <LF> using '\n' as record or line terminator.
However many of the reference files found under ./result have by default a <CR><LF> have a '\r\n' combination as line / record terminator.
(among from additional block characters) and this leads to a miss comparsion and makes the test regressing.
As it looks for me, I have to clean up ./result filesd 1 by 1 and rerun regression test.
Question:
Could it be that during the installation of a UNIX source kit on my PC the ./result files have picked up an additional <CR> '\r' character?
What is the XML standard? How are XML records termiated correctly? by <CR><LF> or just by <LF>? I learned by <LF> only
Which ./result files are of fixed lengtgh records? i.e. 512 bytes long
BTW: libxml2 has build OK with a few warnings only for OpenVMS. I have build it WITH_TRIO and did disable SAX1
How shall I go on with this problem?
If anyone can answer me this few questions that would be very greate. Thanks you in advance.
Regards
Josef.Stadelmann
@axa-winterthur.ch