In D14522#357923, @ostroffjh wrote:

There are several people who are against this, and only one in favor. Would you please consider others' opinions, and defer this further, or put it in a separate branch?

There are several people who are against this, and only one in favor. Would you please consider others' opinions, and defer this further, or put it in a separate branch?

I rebased the patch to fit the latest master branch.
Please test and approve.

Worked here on my openSUSE system.

Other than that it looks good to me.

Looks good to me. We should give it a try and see how the nightly build performs.

Changed source links to our own.

In D14926#311442, @tbaumgart wrote:

Cool, I started out to read about appimage a few times myself but never got around to really work on it. I have not tried any of that yet, but it looks very promising. Can we have that for the upcoming 5.0.2 release?

Cool, I started out to read about appimage a few times myself but never got around to really work on it. I have not tried any of that yet, but it looks very promising. Can we have that for the upcoming 5.0.2 release?

Incorporated changes according to the review.

In D14756#307613, @tbaumgart wrote:

In general, the idea to change the storage of matching transactions is OK. It was more of a quick hack back then. But I think, we need to iron out a few things before we can add it to master. Maybe even postpone it after the next release, because it changes the file structure and is not backward compatible.

In general, the idea to change the storage of matching transactions is OK. It was more of a quick hack back then. But I think, we need to iron out a few things before we can add it to master. Maybe even postpone it after the next release, because it changes the file structure and is not backward compatible.

Looks OK to me. Could not really test it thouroughly, since I don't use the DB backend.

Works for me. Seems to be OK.

In D14522#302502, @tbaumgart wrote:

In D14522#302220, @wojnilowicz wrote:

In D14522#302070, @habacker wrote:

Christian David mentioned at https://mail.kde.org/pipermail/kmymoney-devel/2018-May/020846.html to wait until 23rd of October 2018 before merging alkimia in case no feature has been added.

Who's going to add new features?
That is Christian's point of view. I'm not going to sit and wait longer for a miracle to happen. Nobody develops alkimia and it's unhandy to fix compiler warnings with it being in a separate library.

It is also my POV. Alkimia is stable and does its job. At this point, I don't care for compiler warnings. We certainly have more severe problems than compiler warnings which might have been introduced recently and exist more or less on Windows environments only. I don't want to jeopardize the current stability by changing one of the core elements of KMyMoney. That said, I am against this change to be applied to the current master or 5.0 branch at least until the next KMyMoney release.

In D14522#302220, @wojnilowicz wrote:

In D14522#302070, @habacker wrote:

Christian David mentioned at https://mail.kde.org/pipermail/kmymoney-devel/2018-May/020846.html to wait until 23rd of October 2018 before merging alkimia in case no feature has been added.

Who's going to add new features?
That is Christian's point of view. I'm not going to sit and wait longer for a miracle to happen. Nobody develops alkimia and it's unhandy to fix compiler warnings with it being in a separate library.

In D14522#302070, @habacker wrote:

Christian David mentioned at https://mail.kde.org/pipermail/kmymoney-devel/2018-May/020846.html to wait until 23rd of October 2018 before merging alkimia in case no feature has been added.

Christian David mentioned at https://mail.kde.org/pipermail/kmymoney-devel/2018-May/020846.html to wait until 23rd of October 2018 before merging alkimia in case no feature has been added.

I'm at least somewhat against this change. Maybe it's just wishful thinking, but if we improve Alkimia, maybe there will be other users. That becomes essentially impossible if the code is moved internal to KMM.

Looks OK to me, though I cannot really test it other than looking at the written data.

Looks OK to me. I have checked that my datafile can be loaded and produced identical results as master after writing and reading back. This is the case and the tests work.

Looks OK and if I apply the following patch it also compiles for me (the first hunk seems to be caused by trailing blanks) :

In D14257#295467, @tbaumgart wrote:

Tested with my real file which contains 66 schedules. Produces the exact same output as master.

Tested with my real file which contains 66 schedules. Produces the exact same output as master.

In D14257#295447, @tbaumgart wrote:

I am missing the removal of the old version in MyMoneySchedule::writeXML().

I am missing the removal of the old version in MyMoneySchedule::writeXML().

This package does not match my distro and I honestly don't want to install it. But since I can compile w/o this feature enabled, go ahead and add it.

Looks OK to me. Made some tests in comparing the output of the master version and this version which do not show differences.

I needed to apply the following patch to get correct results. Reason: you must not set the entry date of a scheduled transaction to a valid value, due to the fact that

I think it's not a good idea to improve code on such occasions. Code could be easily broken and it would be hard to find a cause.

Other than that it looks good to me.

In D13817#291800, @tbaumgart wrote:

Now it fails with

[ 98%] Building CXX object kmymoney/plugins/sqlcipher/CMakeFiles/qsqlcipher.dir/qsql_sqlite.cpp.o
/home/thb/devel/kmymoney/build/kmymoney/plugins/sqlcipher/qsql_sqlite.cpp:44:10: fatal error: 
      'QtSql/private/qsqlcachedresult_p.h' file not found
#include <QtSql/private/qsqlcachedresult_p.h>
         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 error generated.

I am not sure if using any of these private headers is a good idea. Where does this file come from? I cannot find it for my distro anywhere (maybe, I have not looked everywhere yet).

Now it fails with

[ 98%] Building CXX object kmymoney/plugins/sqlcipher/CMakeFiles/qsqlcipher.dir/qsql_sqlite.cpp.o
/home/thb/devel/kmymoney/build/kmymoney/plugins/sqlcipher/qsql_sqlite.cpp:44:10: fatal error: 
      'QtSql/private/qsqlcachedresult_p.h' file not found
#include <QtSql/private/qsqlcachedresult_p.h>
         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 error generated.

I am not sure if using any of these private headers is a good idea. Where does this file come from? I cannot find it for my distro anywhere (maybe, I have not looked everywhere yet).

Yes it does.

Does this make any difference?

On my system Qt5::SqlPrivate is defined in Qt5SqlConfig.cmake which is installed by qt5-qtbase-devel. If you won't manage to get Qt5::SqlPrivate on your system, then we'll use less elegant way.

Tried to fix sqlcipher.json issue.

After installing the necessary sqlcipher packages, I get the option to be found and turned on, but cmake produces the following output

I used File/Dump memory with master and D13831 and here is the difference after loading the file

In D13817#290522, @tbaumgart wrote:

The patch compiles now, but the option is disabled here. I don't have all the necessary packages and don't use SQL backends. Can you find someone else who can verify that this change works as expected and does not break anything?

In D13831#290516, @tbaumgart wrote:

In D13831#290468, @wojnilowicz wrote:

In D13831#290049, @tbaumgart wrote:

Did you ever thought about using Q_ENUM? See https://woboq.com/blog/q_enum.html for some details. It looks cleaner if it can be used.

I believe it needs a class with Q_OBJECT and probably would require some more includes in headers, so I don't want to switch to it now.

Ah, that could well be, since it relies on the QMetaObject stuff.

In case I load a file with this patch, the Update all accounts button is grayed out. This works OK on master with the same file. Can you please check all actions once you have fixed this?

The patch compiles now, but the option is disabled here. I don't have all the necessary packages and don't use SQL backends. Can you find someone else who can verify that this change works as expected and does not break anything?

In D13831#290468, @wojnilowicz wrote:

In D13831#290049, @tbaumgart wrote:

Did you ever thought about using Q_ENUM? See https://woboq.com/blog/q_enum.html for some details. It looks cleaner if it can be used.

I believe it needs a class with Q_OBJECT and probably would require some more includes in headers, so I don't want to switch to it now.

In D13831#290049, @tbaumgart wrote:

Did you ever thought about using Q_ENUM? See https://woboq.com/blog/q_enum.html for some details. It looks cleaner if it can be used.

Cannot build w/o sqlcipher being installed. Sqlcipher must be optional not mandatory.

Did you ever thought about using Q_ENUM? See https://woboq.com/blog/q_enum.html for some details. It looks cleaner if it can be used.

It compiles but I spotted a problem in the graph of one of my accounts. I expect it not to be related to this patch which I verified.