Well, the problem is clearly in testinput2 rather than in the program. I did not see it as I, sort of, expected testinput2 to fail and relied more on my EBCDIC oriented subset of testinput2 (which does not contain #loadtables.) I will address that issue in the next version (10.36,) but in any case I think that there should be a way to override the period. Something like:
#loadtables ${basedir}/testdata/testbtables
where ${basedir} is defaulted to '.', but could be overriden. As for the EBCDIC, yes, my build deals with it and generates EBCDIC compatible tables, but it is in the binary library. I guess that this is just a different terminology between you and I. In my lingua source and binaries are separated, even though both may be part of the distro (unless you are compelled to build it from the source if you have a different flavor of EBCDIC.) Also, using the prescribed DFTABLES with option -b, I should probably be able test against ASCII which should be interesting... will try in next release as well.
Ze'ev Atlas
On Sunday, November 22, 2020, 03:53:48 AM EST, Daniel Richard G. <skunk@???> wrote:
Hi Ze'ev,
On Sat, 2020 Nov 21 19:49-05:00, Ze'ev Atlas wrote:
> Hi Phil
>
> I see that this involved an open file command which means that you have
> to supply the filename, potentially fully qualified filename. That
> brings a few questions:
>
> 1. Why should the onus of fixing this be on you rather than on the
> user, who should supply a fully qualified file name?
The relative pathname in question is present in the testdata/testinput2
file. It is not provided by the user.
> 2. The file loaded is in binary format, so it should not be the source
> code directory that is passed to pcre2test but the binary directory,
> unless, of course if it is the same one in your OS and build setting
> (which cannot be in native z/OS)
The "testbtables" file is in binary format, but it is part of the PCRE
source distribution. The file is probably not usable when processing
EBCDIC text, however.
--Daniel
--
Daniel Richard G. || skunk@???
My ASCII-art .sig got a bad case of Times New Roman.
