I have one more piece to add to JENA-992. I have it at home working and
will add it this evening. Following that a clean build will close the
issue.
On Fri, Jul 24, 2015 at 11:41 AM, Rob Vesse <rve...@dotnetrdf.org> wrote:
> Comments inline
>
> On 24/07/2015 10:12, "Andy Seaborne" <a...@apache.org> wrote:
>
> >Rob,
> >
> >Still some problems :-(
> >
> >
> >1/ FREE_MEM
> >
> >For production use, are you expecting warnings on every run when the
> >file size is larger than free memory? I have typically used tdblaoder2
> >on large files way bigger than memory so a WARN each time the program
> >runs is a bit excessive.
>
> Agreed, a couple of warnings have been reduced to debug
>
> >
> >It isn't being calculated correctly - on non-darwin it calls
> >
> >FREE_MEM=$(free -b)
> >
> >and does no further processing but that is a long messy string.
> >
> >Which figure from multi-line output from "free -b" is it looking for?
> >(what's {6} on OSX top output?)
> >
> >
> >echo $OSTYPE
> >==>
> >linux-gnu
>
> This appears to be a 2.x vs 3.x kernel issue
>
> When I tested this on some of our internal Linux servers (which are 2.x
> kernels) free -b just returns an integer, as you note on newer (3.x I
> assume) kernels it instead prints a more complex output
>
> I have changed the logic to now check for complex output and extract the
> desired value and to do an extra check for numerics with errors suppressed
> and to return the not available value if not numeric
>
> I have also restricted the functionality to just OSTYPE linux*
>
> >
> >
> >
> >2/ The use of pv is causing terminal problems (sic).
> >
> >In a gnome terminal, character echo is turned off when tdbloader2 exits.
> >In an emacs shell buffer, the load hangs.
> >
> >Haven't found out why pv is doing this. I wonder if two in the same
> >line are causing character mix up on output leading to broken terminal
> >control sequences, especially with the small file I was using.
> >
> >It's a really nice feature to have to see the progress. Would it be
> >safer for this release to switch off pv to give time for testing in
> >different environments (other *nixes, cygwin) with a simple HAS_PV=0 ?
>
> Agreed, actually it is HAS_PV=1 because I just use the return of `which
> pv` to detect it so 0 is enabled and non-zero is disabled
>
> Have removed the auto-detection and defaulted it to off but users can set
> HAS_PV=0 in their environment if they known pv works reliably in their
> environment
>
> Rob
>
> >
> > Andy
> >
> >On 23/07/15 15:40, Rob Vesse wrote:
> >> Comments inline:
> >>
> >> On 23/07/2015 14:41, "Andy Seaborne" <a...@apache.org> wrote:
> >>
> >>> Trying to do a release, I came across some issues.
> >>>
> >>> JENA-992: (Refactor graph/permissions interface layer)
> >>>
> >>> Not sure of the status of this but I'm assuming that the code already
> >>>in
> >>> 'master' is releasable.
> >>>
> >>> JENA-997: (tdbloader2 script refactoring)
> >>>
> >>> The new scripts misbehave on Linux - there isn't one (obvious) issue.
> >>>
> >>> To unblock the release, if there is a small fix, then great. Another
> >>> possibility is to revert to the older scripts for 3.0.0, so as to fix
> >>> afterwards. This gives more time and space for testing.
> >>
> >> Looks to be relatively simple, think I have the bugs you identified
> >> resolved
> >>
> >> For Case 1 I needed to look up the drive info based on the directory
> >>where
> >> the work files will be created and not the work file itself because that
> >> doesn't exist yet. As part of fixing this I also made the script
> >> resistant to errors where the drive information was unavailable
> >>
> >> For Case 2 I was checking the directory before I had ensured it existed
> >> and was a directory so that just required changing the order of checks
> >>
> >> Rob
> >>
> >>>
> >>> It looks to me like "bash" on OSX is bash 3.2 derived (3.2 was
> >>> originally 2006) whereas on Ubuntu currently it is 4.3. There might be
> >>> other issues that arise if the current ones are resolved with bash or
> >>> other commands.
> >>>
> >>> Andy
> >>
> >>
> >>
> >>
> >
>
>
>
>
>
--
I like: Like Like - The likeliest place on the web
<http://like-like.xenei.com>
LinkedIn: http://www.linkedin.com/in/claudewarren
