Open GoogleCodeExporter opened 9 years ago
[deleted comment]
Hello Daniel,
unfortunately I had to find the above comment seems to be true in more than
just one single case.
I tested a phrase "lpr_gwy" in an encrypted and quite nested file with 2 hits
verified.
It was missed in some older versions of KN, but was fixed by you within the
1.7.x.x family and was working in 1.7.9.3 Beta 3.
In 1.7.9.3 Beta 4 it is missed again completely (0 hits) regardless which
search method was used (Ctrl+F from within the node that contained the phrase
or the F9 search panel) and a bing sound is played with some delay.
The phrase is also not marked in my case if the node is displayed.
System used is Windows 8 32 Bit
I will test it on XP next week.
Original comment by stati...@gmx.net
on 12 Jan 2013 at 12:34
Ok, I will try to reproduce it. Also, I encounter really strange what both of
you tell about a sound played while find is executing ¿?
Thanks
Rikkertbf, the version is 1.7.9.4 or 1.7.9 Beta 4. When named the .rar file I
made a mistake and now I can't change it withour removing and uploading it
again.
Original comment by dpra...@gmail.com
on 18 Jan 2013 at 8:20
[deleted comment]
[deleted comment]
[deleted comment]
Hello Daniel,
I managed to prepare a test file for you. (attached _TestSearchRegression -
cleanedup2.knt)
The encryption has nothing to do with the issue, so I removed it.
The structure reads like:
XXXXX scratch
qms_xxxxx.txt
Software APPS
WAN
OLISNet
config
XXXXX.old
The culprit lies within "config"
If this node is positioned before the node that contains the search result it
prevents a successfull search.
If you move it down to the end of the tree (or if you start the search from
config marked) everything is fine.
In short:
When you search "lpr_gwy" from the top of qms_xxxxx.txt you will get a 0 result.
When you search from config (or move config down the tree) you will get 2
results.
To me it looks like there is a problem with the rtf coding, may be it gets
somewhat unclean over time.
Generally i observed during the cleanup operation that there are still code
chunks from deleted nodes in the rtf source that are not shown in KN.
What makes me wonder is: 1.7.9. Beta 3 has no problems with the same rtf code
whilst 1.7.9.4 obviously has.
Original comment by stati...@gmx.net
on 19 Jan 2013 at 11:30
Attachments:
Fixed at revision r179
Staticip, your test file have been a great help.
The problem was caused by a fragment (a letter) of text included in 'config'
node marked as protected (you can see that RTF appears as: "usual password
xxxxxx\protect x\protect0 x.")
That isn't correct, but the auxiliar edit control used in the find process was
not configured to respond on ProtectChange events and, by default, changes
affecting protected text are discarded. The editors associated to the different
notes are configured to always allow that changes and so you can easily modify
that text without problem. But the auxiliar edit control wasn't, never.
When I began maintaining KeyNote I detected this problem in some rare
ocassions, but I really didn't find why the auxiliar edit control refused to
load another stream of data. The solution I found (brute force but worked...)
was to free and recreate the auxiliar control if, after doing .Clear and before
loading the stream, the control continued having text. In Beta 4 I have done a
complete rewritten of the find/search problem and I forgot to include it.
With your example I have found easily the cause of the problem and so, the
solution have been the correct one.
(The 'bip' sound you mentioned was raised by the edit control when refusing
changes in protected text, one for each attempt (4 times, four nodes after
'config'))
Regards
Original comment by dpra...@gmail.com
on 26 Jan 2013 at 11:00
Thanks, you make me very happy (I hope I can get my dirty fingers on the next
compiled clean version soon ;-) since I like the reworked optic on the search
panels result list very much)
As I do not know of any option within KeyNote to set a protection tag I
researched a little in the meantime and probably found the origin for this
problem.
the \protect ... \protect0 tag must have come into KN via a pasted e-mail from
Outlook that contained this tag already.
I scanned my knt files meanwhile and the test part I sent to you was the only
one containing this tags.
So in theory one could also manually parse the document from time to time to
find and remove the offending tags.
But I like your brute force solution more as it is automatic.
KN seems not the only software that has to fight with this problem, there are
numerous posts around that show that the \protect tag can cause problems in all
sorts of places.
just one example http://www.devexpress.com/Support/Center/p/B147444.aspx
...
About the protected tag: I am NOT USING IT, I am just allowing my users to
paste any rtf text they like, they can either type, or copy and paste from
another source. In the particular case of my unhappy customer he copies and
pastes an e-mail text from MS Outlook, and that rtf text contains "\protect".
...
same scenario
thanks for the fix (I'm sure this is also the source of search problem that KN
1.6x was suffering from).
Original comment by stati...@gmx.net
on 26 Jan 2013 at 4:06
Hello,
The final solution has not been the old 'brute force' but something simpler,
just to listen to ProtectedChange event and respond allowing it ;)
Best regards
Daniel
Original comment by dpra...@gmail.com
on 26 Jan 2013 at 4:16
Original issue reported on code.google.com by
rikker...@gmail.com
on 10 Jan 2013 at 11:41