On 2017.09.18 14:46, Thomas Baumgart wrote:
On Sonntag, 17. September 2017 16:13:49 CEST Jack wrote:
> On 2017.09.17 02:40, Thomas Baumgart wrote:
> > On Samstag, 16. September 2017 17:26:06 CEST Jack wrote:
Having solved the libalkimia problem (or perhaps non-problem) I
successfully compiled from git-head master branch. It is
certainly better than the last time I tried it, probably months
ago since I got a clean compile. Hopefully we are really getting
close to at least a pre-release release, to address those distros
which no longer have KDE4 pieces around to rely on.
> >>
> > > However, I have two issues.
> >>
First, whenever I display a ledger, I get lots of "Invalid sort
key -9" where the number seems to alternate between -4 and -9, and
I get dozens to hundreds of copies. It doesn't seem to hurt
anything, and the sort order of the ledger appear correct.
> >
Then you don't have git master, since I fixed this up in Randa.
Seeing this, the sorting in the ledger was broken (if descending
was selected).
>
Hmm. I was sure I had just done a git pull before compiling, but it
seems you are right. A new git pull and redo everything and no more
sort errors.
> [....]
>
Second, is the ClientUid for libOFX udpates actually enabled?
When I try to update an account which was mapped using a recent
4.8 git version to successfully use this, I get that stupid error
from Chase, and ofxlog.txt (once I figured out how to re-enable
it) does not show KMM sending this value. In addition, when I
edit the account, and look at the online data, it does show this
value, but it is displayed in very pale gray, and is not a
selectable field. Is there a problem, or is it just not fully
converted to KF5 yet?
> >
It should be auto-detected during cmake and the result of this
detection shows up in the cmake summary. If can restart the scan,
by removing <your-build- directory>/CMakeCache.txt and rerun cmake
with all the options you passed. The summary should show, if the
ClientUid option is available or not. In case that does not work,
we need to fix it (my distro version of libofx (0.9.10) does not
support it, so it's off for me by default).
>
I'm still not sure about this one. I have libofx 0.9.11 instaled,
and I can find no remnants of any other version. My system version
of KMM 4.8.0 does not know about ClientUid - the edit account
doesn't show the clientuid at all 4.8 compiled from git head DOES
know, displaying it on one of the tabs of the online detail section
for teh account. git master says it knows about clientuid in the
cmake summary, but although the clientuid is displayed in the edit
account dialog, it is disabled/grayed out.
>
> [snip all my ramblings about possible causes...]
>
> Any other thoughts on where I might look for relevant differences?
I just checked that. I have a libofx version w/o the clientuid and a
build environment setup which created a KMyMoney version with OFX
support, but w/o the clientuid.
Then I simply changed the libofx.h header file to contain the resp.
field. Simply running cmake again does not turn the clientuid on. I
had to remove my CMakeCache.txt file and rerun the complete cmake
from scratch to pick that up. So it seems, that some of that stuff is
cached by cmake.
Yes, I know cmake caches everything that it does find.
The ONLY version of libofx on my system (0.9.11) has the clientuid
peces. When I did cmake or ccmake, the summary said clientuid support
IS included. When I run KMyMoney, the Client UID is displayed on the
OFX Details tab of the Online settings tab of the account edit dialog,
but grayed out, and it is not sent as part of the OFX request.
After deleting CMakeCache.txt and running cmake again, I do get
OFX plugin: yes
OFX supports CLIENTUID: yes
as part of the summary. Doing make and make install seems to have
relinkned many of the .so files, but the behavior is the same. I'll
try a make clean and another make, but I don't think it will change the
behavior. I find this rather confusing.
Jack
