Feed Advanced Search

Yesterday

mwolff added a comment to D8852: Fix crash deep inside VcsOverlayProxyModel::data().

lgtm too, does it noticeably influence startup time maybe?

Thu, Nov 16, 8:54 PM
mwolff created D8850: WIP: support drag and drop between shared folder view containments.
Thu, Nov 16, 4:15 PM · Plasma
mwolff updated the diff for D8598: FolderView: position files at drop event target position.

update message

Thu, Nov 16, 2:14 PM · Plasma
mwolff updated the diff for D8598: FolderView: position files at drop event target position.

use filename in mapping to support desktop:/

Thu, Nov 16, 2:12 PM · Plasma
mwolff added a comment to D8493: Make Folder View screen aware.
[14:24] <kbroulik> andris: milian umm, how do I actually move files between screens?
[14:26] <kbroulik> the fact that you can actually resize it with Alt+right-click is a bug
[14:27] <milian> kbroulik: dnd, which is WIP
[14:28] <kbroulik> milian: okay because right now the screen-aware FV does not work at all. other than that I dont have any icons on my second screen anymore :)
[14:28] <kbroulik> you mean the "position files at drop event"?
[14:28] <milian> yes, but that is only one step
[14:29] <milian> I'm working on the second, more important one
[14:29] <milian> can you tell me what you mean by no icons on the second screen?
[14:29] <milian> did you have some on the second screen and by applying the patch they got moved to the first screen?
[14:29] <milian> i.e. some kind of config porting?
[14:29] <kbroulik> yes
[14:29] <milian> andris: ^ that is useful important feedback, we need to handle that somehow
[14:29] <milian> but kbroulik - how did you have icons on your second screen before?
[14:29] <kbroulik> milian: I have a FV on both screens, both pointing to desktop:/
[14:29] <milian> did you use the same folder?
[14:30] <kbroulik> yes
[14:30] <milian> then both used to show the same icons, no?
[14:30] <kbroulik> they don't
[14:30] <kbroulik> err
[14:30] <kbroulik> yes, they *used to*
[14:30] <kbroulik> so this is somewhat to be expected I guess
[14:30] <milian> right so nothing to be ported after all
[14:30] <kbroulik> yans: well, that's how it is
[14:30] <yans> So i just asking, kbroulik thx for info.
[14:30] <milian> what's missing is the DND to allow people to move files actually
[14:30] <milian> I'm working on that, keep getting sidetracked by other projects
[14:30] <milian> let me sit down on it again
[14:35] <kbroulik> milian: ah, okay, because right now DND is unchanged, ie. I drag files over and it says "would overwrite" :)
[14:36] <milian> yes that is broken as before
[14:36] <milian> I'm working on that and have a local WIP for that though
[14:36] <kbroulik> oki :)
[14:36] <kbroulik> so I cannot really comment on D8493 let's have Sho_ decide
Thu, Nov 16, 1:38 PM · Plasma
mwolff accepted D8493: Make Folder View screen aware.

lgtm from my side. Anyone from the plasma team care to chime in?

Thu, Nov 16, 1:06 PM · Plasma
mwolff added a comment to D7995: KDevelop: address dirwatching inefficiency (WIP/PoC).
In D7995#168405, @rjvbb wrote:

Milian Wolff wrote on 20171116::10:55:31 re: "D7995: KDevelop: address dirwatching inefficiency (WIP/PoC)"

I **still** think the number of watched directories is completely useless information. The total time is interesting, and whether the unit tests still work that ensure the correct dirs are watched. This means: remove the project watcher class, keep using dirwatcher directly.

We still disagree on that subject then and I can still keep the benchmark crippled if that's what you want. Did you actually read the argument why I reintroduced this output, how total time is now dependent on a number of watched directories that is no longer an external given but is now essentially an *un*controlled variable.

Consider 2 directories with exactly the same number of files and directories. One has a substantial subtree under a folder that will be filtered out, the other not. Which one do you think will import faster, and do you think you'll understand that difference at a glance from benchmark output that doesn't includes total time but not the number of timed operations?

Thu, Nov 16, 12:22 PM · KDevelop
mwolff requested changes to D8493: Make Folder View screen aware.

some more nits, sorry for that ;-)

Thu, Nov 16, 12:13 PM · Plasma
mwolff requested changes to D7995: KDevelop: address dirwatching inefficiency (WIP/PoC).

I still think the number of watched directories is completely useless information. The total time is interesting, and whether the unit tests still work that ensure the correct dirs are watched. This means: remove the project watcher class, keep using dirwatcher directly.

Thu, Nov 16, 10:55 AM · KDevelop
mwolff added inline comments to D8493: Make Folder View screen aware.
Thu, Nov 16, 10:46 AM · Plasma

Wed, Nov 15

mwolff added a comment to D8332: Added baloo urls into places model.

lgtm, one minor nit, potentially for the future

Wed, Nov 15, 2:43 PM · Frameworks
mwolff accepted D8348: Add a section for removable devices.

lgtm too

Wed, Nov 15, 2:37 PM · Frameworks
mwolff requested changes to D8450: User can now hide an entire places group from KFilePlacesView.

please use arc, or add more context to your patches in the future

Wed, Nov 15, 11:06 AM · Frameworks
mwolff requested changes to D8367: Hidding place groups implementation in KFilePlacesModel.
Wed, Nov 15, 10:56 AM · Frameworks
mwolff added inline comments to D8348: Add a section for removable devices.
Wed, Nov 15, 10:36 AM · Frameworks
mwolff requested changes to D8348: Add a section for removable devices.

the UDI comment needs to be investigated, otherwise lgtm

Wed, Nov 15, 10:35 AM · Frameworks
mwolff committed R32:8b37c39dd327: Merge branch '5.2' (authored by mwolff).
Merge branch '5.2'
Wed, Nov 15, 10:27 AM
mwolff committed R32:da49f7bf76aa: fix "none" being set as default compiler on windows (authored by geetamc).
fix "none" being set as default compiler on windows
Wed, Nov 15, 10:27 AM
mwolff closed D8805: fix "none" being set as default compiler on windows.
Wed, Nov 15, 10:27 AM
mwolff accepted D8805: fix "none" being set as default compiler on windows.

lgtm, will push to 5.2

Wed, Nov 15, 10:27 AM
mwolff added inline comments to D8642: Rework saving of annotations and form data.
Wed, Nov 15, 10:22 AM · Okular
mwolff added a comment to D8348: Add a section for removable devices.

Could you add camera:/ devices to this section too? That way we could also take care of https://bugs.kde.org/show_bug.cgi?id=386060

Wed, Nov 15, 10:19 AM · Frameworks
mwolff added a comment to D8348: Add a section for removable devices.

What about something like this as well

The removable device gets an "unmount" or "disconnect" icon on hover?

Thoughts?

Yes yes yes! That would also resolve https://bugs.kde.org/show_bug.cgi?id=154499

Of course, if this isn't an easy change, we may not want to overload the patch.

Wed, Nov 15, 10:19 AM · Frameworks
mwolff added a comment to D8642: Rework saving of annotations and form data.

I started with a cursory glance at the changes, but the change set is huge, which makes it really hard to review, especially for people who have no extensive prior experience with okular. I think some of my comments could be "solved" by adding a comment here or there, as it seems like they are non-issues but rather arise from misunderstanding or lack of information from my side. Help future people looking at this code by writing some more comments. And consider splitting up such monster commits in the future into smaller patches, one building on the other - if possible. Or is this the case already? I.e. could I instead review a feature branch? I know that phabricator sucks in that regard, but at least it would simplify my life at reviewing, even if I'd read patches on the command line and then add comments here

Wed, Nov 15, 9:59 AM · Okular
mwolff added inline comments to D8493: Make Folder View screen aware.
Wed, Nov 15, 9:25 AM · Plasma
mwolff committed R119:b985b7432ef9: Cleanup: use temp variables instead of querying properties repeatedly (authored by mwolff).
Cleanup: use temp variables instead of querying properties repeatedly
Wed, Nov 15, 9:06 AM
mwolff closed D8597: Cleanup: use temp variables instead of querying properties repeatedly.
Wed, Nov 15, 9:06 AM · Plasma
mwolff added a comment to D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.

yes, only bug fixes go to stable branches

Wed, Nov 15, 8:57 AM · KDevelop

Tue, Nov 14

mwolff accepted D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.

thanks, lgtm now

Tue, Nov 14, 9:47 AM · KDevelop
mwolff requested changes to D8805: fix "none" being set as default compiler on windows.

idea lgtm in general, but please cleanup the implementation

Tue, Nov 14, 9:46 AM

Mon, Nov 13

mwolff requested changes to D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.

instead of using a friend, do what the other tests do - inherit from the ProjectController, make addProject public, then set your custom controller on the core.

Mon, Nov 13, 1:20 PM · KDevelop
mwolff added a comment to D8654: Don't ask for compiler-specific includes in QmlJS.

Imo this makes the code more consistent, as you say. So feel free to push it, too.

Mon, Nov 13, 1:07 PM

Mon, Nov 6

mwolff accepted D8653: Ignore unrecognized files in CompilerProvider.

lgtm, but you can simplify the code by writing return {}; in both places. Also, it doesn't hurt to add a short comment saying something like you do in your commit message

Mon, Nov 6, 12:28 PM
mwolff accepted D8654: Don't ask for compiler-specific includes in QmlJS.

lgtm

Mon, Nov 6, 12:27 PM

Sat, Nov 4

mwolff accepted D8644: Recognize MIME type text/x-opencl-src as OpenCL.
Sat, Nov 4, 7:32 PM

Fri, Nov 3

mwolff accepted D8638: Remove DataModel::roleNameToId.
Fri, Nov 3, 11:20 AM · Frameworks, Plasma
mwolff accepted D8639: Optimize SortFilterModel role names.
Fri, Nov 3, 11:19 AM · Frameworks, Plasma
mwolff added a comment to D8634: Custom defines/includes: Improve handling of parser arguments.

Overall I'm in favor of where this is heading. I've added a bunch of comments, but then I figured - couldn't we simply replace ParserArguments with e.g. a QHash<Utils::LanguageType, QString>? Or, better yet, we make it a QString[Utils::LanguageType::Other]? That would simplify a lot of the code you are changing/adding, while still enabling your approach. The advantage is that these types are easily iteratable and thus isAnyEmpty becomes a simple std::any_of all, the .get() and .set() calls get replaced by operator[] calls, and you'd still be able to use brace-initialization, while having to write less code overall.

Fri, Nov 3, 10:50 AM
mwolff accepted D8588: Make VCS data container classes implicitly shared.

lgtm. and to be clear: better verbose tests than no tests. so thanks! when i have time i'll try to cleanup the tests a bit to show how i think one could reduce the verbosity

Fri, Nov 3, 10:17 AM

Thu, Nov 2

mwolff added a comment to D8059: KDevelop: abstractfilemanagerimport benchmark.
In D8059#160968, @rjvbb wrote:

So, where do we go from here?

I started preparing the introduction of the ProjectWatcher class (which was rejected from the benchmark introduction patch) but then realised that it isn't justified as a standalone commit. That class only makes sense if you're going to use it other than by handing the project top dir to ProjectWatcher::addDir, telling it to add everything under that directory.

The class makes sense as part of my patch that lets FileManagerListJob determine which directories (dirs only!) get watched, and it's with that use pattern that it makes sense to check if the directory trips a filter.

IOW, I'm not in favour of going through a review just to introduce a ProjectWatcher class knowing it does things (like locking a mutex) that are superfluous in the current context.

As to checking the project filter in ProjectWatcher::addDir() : it seems obvious that this *must* have a significant benefit under the right conditions (in-tree build dir and/or the .git directory for projects with a matching exclusion filter). Is it really necessary to go through an additional review just to add the filter check?

Thu, Nov 2, 9:48 PM · KDevelop
mwolff requested changes to D8493: Make Folder View screen aware.
Thu, Nov 2, 5:35 PM · Plasma
mwolff added a comment to D8379: PDF: Support the new poppler renderToImage with update callback.
In D8379#163631, @rkflx wrote:

@aacid In general I notice the changes you are making, sorry I'm not always able test immediately. I'll try to have another look in the next couple of days.

(That said, more people testing by actually running the code with the newest Poppler would be great.)

Thu, Nov 2, 5:33 PM · Okular
mwolff added a comment to D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.
In D8579#163619, @rjvbb wrote:

So I think you simply don't have a clue of how filtering is operating.

Indeed, and I'd just as well remain blissfully unaware of its internal cuisine :)

Maybe they occur less often for you. I never open folders that contain any `*.o`, since I always build out-of-source. So clearly haggling over order is a waste of time.

Ah, because you're the standard and shining example and we should all be your clones? Let me repeat here that I've never been able to set up a usable KDevelop project for an autoconf-based project using an out-of-source build directory.

Thu, Nov 2, 5:06 PM · KDevelop
mwolff added a comment to D8588: Make VCS data container classes implicitly shared.

Big change is big. Why did you bother doing it? Do you have performance measurements indicating that this is a bottleneck? If so, do you want to save memory or do you want to make things faster? Would it maybe be enough to add move semantics?

Thu, Nov 2, 4:17 PM
mwolff added a comment to D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.
In D8579#163574, @rjvbb wrote:
I once again fail to understand you. What is a "properly set up filter" vs. what you are measuring here? Both should do the same thing now, no? I.e. open the project and load the configuration that contains custom filters to exclude e.g. the build dir.
>   >   Also, how would reordering default filters change anything?

Hmm, you mean it doesn't really matter if you first check a directory called "build" against all "*.a", "*.o" and whatever uninteresting files that directory might contain before coming across the rule that excludes the entire directory, because either way you won't be scanning it?

Thu, Nov 2, 4:06 PM · KDevelop
mwolff added a comment to D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.
In D8579#163409, @rjvbb wrote:

thanks for the numbers and profile on the project filter, I guess I'll need to port it away from QRegExp to QRegularExpression, which will mean some custom parser to transfer unix style matching to regular expressions patterns. But it does seem to help (left is status quo, right is QRegularExpression with manual conversion of the default patters to regular expressions):

...

Anyhow, not really related to this patch here, but good to keep in mind for the future

Indeed. Are those gains still visible in real world usage - it doesn't look like filtering is a bottleneck operation like the dirwatcher population currently is? In a real world project import, a properly set up filter would exclude an in-tree build directory first, thereby skipping a whole bunch of files that would otherwise be excluded one by one. (IOW, it might already be very effective to optimise the order of the entries in the default filter.)

Thu, Nov 2, 2:34 PM · KDevelop
mwolff retitled D8598: FolderView: position files at drop event target position from WIP: position files at drop target position in folder to FolderView: position files at drop event target position.
Thu, Nov 2, 2:07 PM · Plasma
mwolff updated the diff for D8598: FolderView: position files at drop event target position.

use copyjob signals to map the target urls, not the source urls, which
resolves a lot of the brittleness of my previous approach. overall, this
works quite nicely, imo and is already in an acceptable state (to me).

Thu, Nov 2, 2:01 PM · Plasma
mwolff added inline comments to D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.
Thu, Nov 2, 1:59 PM · KDevelop
mwolff accepted D8379: PDF: Support the new poppler renderToImage with update callback.

lgtm from my side

Thu, Nov 2, 12:09 PM · Okular
mwolff added inline comments to D8379: PDF: Support the new poppler renderToImage with update callback.
Thu, Nov 2, 11:03 AM · Okular
mwolff added a comment to D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.

thanks for the numbers and profile on the project filter, I guess I'll need to port it away from QRegExp to QRegularExpression, which will mean some custom parser to transfer unix style matching to regular expressions patterns. But it does seem to help (left is status quo, right is QRegularExpression with manual conversion of the default patters to regular expressions):

Thu, Nov 2, 10:20 AM · KDevelop
mwolff abandoned D8596: Fix memory leak in FolderModel::drop.

oh of course, the dropJob is the sender, stupid me.

Thu, Nov 2, 9:57 AM · Plasma

Wed, Nov 1

mwolff added a comment to D8598: FolderView: position files at drop event target position.

So far, mostly food for thought. It kinda works but I see lots of problems with this approach... I'll investigate other approaches tomorrow

Wed, Nov 1, 4:52 PM · Plasma
mwolff added a reviewer for D8598: FolderView: position files at drop event target position: amantia.
Wed, Nov 1, 4:51 PM · Plasma
mwolff added a reviewer for D8598: FolderView: position files at drop event target position: hein.
Wed, Nov 1, 4:51 PM · Plasma
mwolff created D8598: FolderView: position files at drop event target position.
Wed, Nov 1, 4:51 PM · Plasma
mwolff created D8597: Cleanup: use temp variables instead of querying properties repeatedly.
Wed, Nov 1, 4:50 PM · Plasma
mwolff created D8596: Fix memory leak in FolderModel::drop.
Wed, Nov 1, 4:50 PM · Plasma
mwolff added a comment to D8493: Make Folder View screen aware.

small nits, plus the asserts from the previous review. otherwise lgtm - the plasma people should chime in

Wed, Nov 1, 3:28 PM · Plasma
mwolff added inline comments to D8493: Make Folder View screen aware.
Wed, Nov 1, 3:24 PM · Plasma
mwolff added a comment to D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.

Oh and do check where the time is being spent - is that really just the filter checks? Do you run your timings on a kdevelop codebase compiled with compiler optimizations? Most notably, is your Qt built with compiler optimizations? My gut feeling is that the filters shouldn't be that bad, performance wise. Again, a profiler will tell you.

Wed, Nov 1, 3:22 PM · KDevelop
mwolff added a comment to D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.
In D8579#162838, @rjvbb wrote:

And always keep the project filter plugin activated?

Wed, Nov 1, 3:14 PM · KDevelop
mwolff added a comment to D8567: Emit signals when a screen is added or removed.

lgtm, but the plasma people should chime in now, esp. regarding what @apol said

Wed, Nov 1, 2:26 PM · Plasma
mwolff added inline comments to D8493: Make Folder View screen aware.
Wed, Nov 1, 2:08 PM · Plasma
mwolff accepted D8434: Created 'remote' section.

lgtm, but please wait a bit, maybe someone else wants to chime in?

Wed, Nov 1, 1:50 PM · Frameworks
mwolff added inline comments to D8434: Created 'remote' section.
Wed, Nov 1, 1:30 PM · Frameworks
mwolff added a comment to D8579: KDevelop: project filter support in the abstractfilemanagerplugin import benchmark.

I'd say always make this load in GUI mode, and remove the switch.

Wed, Nov 1, 1:24 PM · KDevelop
mwolff added a comment to D8434: Created 'remote' section.

But thinking 5s more about it, I personally would say that using "Remote" for the category would be OK for now.

Wed, Nov 1, 12:33 PM · Frameworks
mwolff added a comment to D8434: Created 'remote' section.

Well, but if we use Network for remote:// already, then the group should also have this label, no? I don't see an issue with this, really. On the contrary - maybe we could in the future remove the remote:// link and let the category header react to a click, such that it will show you e.g. all remote devices. That would remove the duplication then.

Wed, Nov 1, 12:31 PM · Frameworks
mwolff added a comment to D8493: Make Folder View screen aware.

some code review from my side

Wed, Nov 1, 11:03 AM · Plasma
mwolff added inline comments to D8567: Emit signals when a screen is added or removed.
Wed, Nov 1, 11:02 AM · Plasma
mwolff added inline comments to D8567: Emit signals when a screen is added or removed.
Wed, Nov 1, 10:41 AM · Plasma
mwolff added inline comments to D8379: PDF: Support the new poppler renderToImage with update callback.
Wed, Nov 1, 10:26 AM · Okular
mwolff added a comment to D8434: Created 'remote' section.

lgtm overall. but I wonder about the naming choice. "Shared" is confusing, to me personally at least. Why not call it "Remote" or "Network"? The reasoning is that "shared" for me is only LAN/WLAN shared stuff, while "remote://" allows arbitrary remote links. I use it to connect via (S)FTP to machines on the internet e.g.

Wed, Nov 1, 10:16 AM · Frameworks

Thu, Oct 26

mwolff accepted D8059: KDevelop: abstractfilemanagerimport benchmark.

please fix the nit, then feel free to commit to master branch. Note again that the commit message should include your measurements you posted above.

Thu, Oct 26, 9:15 AM · KDevelop

Wed, Oct 25

mwolff added a comment to D8465: ItemSync: speed up by not using takeFirst()..

Oh and I notice: The patch doesn't include info about the state after - did you measure it and verify that this is indeed faster? I.e. is the move actually happening and everything faster now?

Wed, Oct 25, 5:12 PM · KDE PIM
mwolff added a comment to D8465: ItemSync: speed up by not using takeFirst()..

lgtm conceptually, but someone else should check that

Wed, Oct 25, 1:08 PM · KDE PIM

Tue, Oct 24

mwolff added a comment to D5425: Allow disabling the git plugin with a kdevgit.disabled git config option.

It completely invisible to new users, I doubt it would be used much. Maybe instead we should honor the bash settings, if the issue is that we are running stat in the background? If the user explicitly runs a git command, we'd still block but that's OK as he asked for it?

Tue, Oct 24, 10:24 AM
mwolff requested changes to D8059: KDevelop: abstractfilemanagerimport benchmark.
Tue, Oct 24, 8:36 AM · KDevelop

Mon, Oct 23

mwolff added a comment to D8059: KDevelop: abstractfilemanagerimport benchmark.
In D8059#159018, @rjvbb wrote:
The number of calls to addDir is totally uninteresting. Even more so as before it was/is recursive internally while you are about to change it. So what value is there when the benchmark shows "1 call to addDir" now, and then eventually "1000 calls to addDir"?!

Again: We only care about the time it takes to import the project.

Please read me carefully. I DO care about more than time alone. If it takes 10 hours to add a quadrizillion directories I might be happy, but not if it takes 10 minutes to add only 5 directories (directory content shouldn't matter since we're watching only directories).

Printing the number of directories loaded is only appropriate if we know that it corresponds to the exact number of directories watched. We don't know that here, there are 2 different models at play.

Please don't try to teach me how to set up experiments (and I won't try to explain what the PS in FLOPS stand for).

Mon, Oct 23, 7:48 PM · KDevelop
mwolff added a comment to D8059: KDevelop: abstractfilemanagerimport benchmark.
In D8059#157415, @rjvbb wrote:

Anyone interested: here's an observation related to benchmarking dirwatching:

I have a session with a project for audacity (2.2.0-rc1, git working copy). I notice that KDevelop becomes noticeably busy on the UI thread when I run configure on that project (in a regular in-tree set-up). I'd be curious to know if others see this as well (with stock KDevelop) - be it with audacity or any comparably large project that contains sources for more than a few 3rd-party dependencies that all use autoconf.

Mon, Oct 23, 6:16 PM · KDevelop
mwolff added a comment to D8059: KDevelop: abstractfilemanagerimport benchmark.
In D8059#157312, @rjvbb wrote:
Note that the number of calls to `addDir` is completely useless output (it's an implementation detail). What is interesting, is the total time only.

I agree that this is true for the current implementation. However, it's the only thing we can measure at the moment without burdening the benchmark or the importer itself with code to count the number of directories in a project. That number of calls should be seen in the light of before and after comparison. I'm fine with meto output this only in a benchmark mod included with the actual dirwatcher modification (in which case it can be printed as "number of dirs watched" since addDir() will be adding a single dir only).

Mon, Oct 23, 6:15 PM · KDevelop

Sun, Oct 22

mwolff accepted D8410: Restore NavigationToolTips for file items in projects manager.

lgtm

Sun, Oct 22, 7:53 AM

Thu, Oct 19

mwolff added a comment to D7995: KDevelop: address dirwatching inefficiency (WIP/PoC).

Note: I'll refrain from reviewing this until we finished with D8059 and the spin-off that then introduces the project watcher to apply the filter early. Once we have those two patches in, we can assess this patch here properly.

Thu, Oct 19, 12:30 PM · KDevelop
mwolff requested changes to D8059: KDevelop: abstractfilemanagerimport benchmark.

the commit message should include some output numbers and document how to run this test with different backends. I.e. something like the following should be the desired commit message _and_ also shows the desired output from the benchmark itself:

Thu, Oct 19, 12:23 PM · KDevelop
mwolff requested changes to D7930: Kdevelop CMake plugin : use canonical paths to build.dir.

No, it should not just be a test around QFileInfo::canonicalFilePath, that is done already in Qt upstream for us. What I want is test(s) for the CMakeManager code you are changing. I.e. take CMakeManager::buildDirectory as an example. Extend the test_cmakemanager (i.e. add another test method there). Then play with symlink structure in a QTemporaryDir to replicate your erroneous setup. Finally call CMakeManager::buildDirectory and ensure it returns "the right thing". Similarly you can call the cmakeutils functions and check their return values for sanity.

Thu, Oct 19, 12:00 PM · KDevelop

Oct 18 2017

mwolff accepted D8199: Allow abortion of starting another NativeAppJob.

ok, lgtm

Oct 18 2017, 7:06 PM
mwolff committed R32:9ec89cdc6a1f: Don't try to guess the language from given arguments (authored by aaronpuchert).
Don't try to guess the language from given arguments
Oct 18 2017, 6:05 PM
mwolff closed D8317: Don't try to guess the language from given arguments.
Oct 18 2017, 6:05 PM

Oct 17 2017

mwolff added a comment to D8317: Don't try to guess the language from given arguments.

what is your email address, that we can use to associate this commit with you?

Oct 17 2017, 7:49 PM

Oct 16 2017

mwolff accepted D8156: Fix registration of the enabled state of plugins.

minor nits, otherwise lgtm

Oct 16 2017, 6:03 PM
mwolff accepted D8317: Don't try to guess the language from given arguments.

lgtm

Oct 16 2017, 8:21 AM

Oct 9 2017

mwolff added a comment to D7930: Kdevelop CMake plugin : use canonical paths to build.dir.
In D7930#153476, @rjvbb wrote:
The test should replicate the situation and show it fail before this patch and then work with this patch. Why is this not possible to do?

This would mean one would have to write a test that

  • sets up full-blown KDevelop session containing a project that's known to fail under the conditions to test and configure it for those conditions (meaning kdevelop itself)
  • launch the session
  • detect if parsing fails
Oct 9 2017, 8:43 PM · KDevelop
mwolff added a comment to D8199: Allow abortion of starting another NativeAppJob.

looking at the old code, I wonder why it did what it did... Most notably, it refetched the current jobs after the message box. This looks intentional and is now gone. I mean sure, while a dialog is shown, anything could happen, including running another job... Do we need to take care of this? Does anyone remember why it was done in this way? Does the git history show up anything?

Oct 9 2017, 8:40 PM
mwolff added inline comments to D8156: Fix registration of the enabled state of plugins.
Oct 9 2017, 7:47 PM

Oct 8 2017

mwolff requested changes to D8156: Fix registration of the enabled state of plugins.

awesome, thanks

Oct 8 2017, 7:26 PM
mwolff added a comment to D7930: Kdevelop CMake plugin : use canonical paths to build.dir.

The test should replicate the situation and show it fail before this patch and then work with this patch. Why is this not possible to do?

Oct 8 2017, 7:18 PM · KDevelop
mwolff requested changes to D8199: Allow abortion of starting another NativeAppJob.
Oct 8 2017, 7:09 PM
mwolff requested changes to D8202: Fix Qt 5.5 compatibility -- no constFirst/Last methods..

I'd actually say let's do the opposite: Let's require 5.6 and drop direct support for the older Ubuntu system. If people want to use a newer KDevelop there, they can simply use the AppImage and be done with it.

Oct 8 2017, 7:02 PM