Re: Cppcheck: Reduction of False Positives: Manual Approach
> It was mentioned that the correct way is to generate the proper project > structure for cppcheck using our existing tools that generate IDE > integrations (and they have correct include correct includes, deps and all). Mike, Yes, agreed. But AFAIK, no one other than me is looking into this. Quite a few issues with the GbuildToIDE approach need to be solved. In the meantime, our current cppcheck script generates tens of thousands of config errors (if run with check-config or verbose),reports hundred or possibly thousands of false positives, and is being called with depreciated parameters. These issues can be fixed now. I am in the process of going through past cppcheck related commits to verify how many, if any at all, would be missed had we been using the '-I include/' parameter. So far, I have not identified any. If the consensus is that you do want to improve the script with the manual approach, I will not continue testing. Please advise. However, keep in mind that from the GbuildToJson output, any framework that is built in the future will also call cppcheck with the '-I include/' parameter. Therefore, if there is a bug with cppcheck not reporting valid errors when called this way, then that approach too will suffer too. ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hi Luke, I believe that continuing the "manual approach" is moot. It was mentioned that the correct way is to generate the proper project structure for cppcheck using our existing tools that generate IDE integrations (and they have correct include correct includes, deps and all). -- Best regards, Mike Kaganski ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
> Was it not you how came up with the idea to reduce the false positives with > specifying the includes? No, it was not my idea. On #cppcheck, I was told by danmar, the primary developer of cppcheck, that our script is using cppcheck incorrectly. Without being passed the same include locations as we pass the compiler, we should expect a large amount of garbage. In fact, according to the developer, we should not get any False Postives if we call cppcheck correctly. He encouraged me to file bug reports for any FP that remain, once cppcheck is being run properly. > The main point that this change seems to simply reduce the scope of cppcheck. > If this is the purpose then we can just run cppcheck on an empty file and so > we won't see any issue (all false positives will disappear). Again, No my goal is to improve the Signal-to-noise. FPs can be dangerous as in tdf#96089 and make it much harder to spot real issues. Currently, I am in the process of comparing old cppcheck fixes with and without the '-Iinclude' option. So far, the three that I have checked would not be filtered out. In other words, had we been calling cppcheck the way I propose, these issues would have been much easier for developers to spot(4000 vs 500). -Luke ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hi Luke, all It does not matter that variableScope is a low priority issue or a can be dangerous . If someone want to disable a check it can be disabled explictiely (I guess). The main point that this change seems to simply reduce the scope of cppcheck. If this is the purpose then we can just run cppcheck on an empty file and so we won't see any issue (all false positives will disappear). >Overall, does this report have a higher signal-to-noise ratio than our >current weekly report? @Luke You do ask it? Was it not you how came up with the idea to reduce the false positives with specifying the includes? Now you are asking others what the result of your change. I'm sure it used to happen in the opposite direction. It does not make any sense to have two reports if the only difference is that first report contains less bugs than the other, if the filtering has no logical meaning. Best Regards, Tamás ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hi Luke, Am October 25, 2018 12:40:33 AM UTC schrieb Luke Benes : >In my first attempt to improve the quality of the cppcheck reports, >Tamás Zolnai pointed out that including every possible header resulted >in some valid warnings not being reported. [snip] >It seems many valid variableScope warnings are still being omitted. I'm >still looking into that. Are there any other categories of valid errors >that are missing? Specific examples would be helpful. > >Overall, does this report have a higher signal-to-noise ratio than our >current weekly report? No idea, you're the expert here - probably that's easier to evaluate in comparison to your original, even shorter report. Why not have two reports? If your final report has much less false positives and probably even generally with a higher error severity (variableScope normally doesn't indicate an error, can it?), then people can concentrate on these first. Now I don't know how long it takes to generate them, but one can still toggle between them. Jan-Marek ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
On 25.10.18 02:40, Luke Benes wrote: It seems many valid variableScope warnings are still being omitted. those warnings are quite dangerous anyway if naively believed, tdf#96089 was quite a pain to debug... ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
In my first attempt to improve the quality of the cppcheck reports, Tamás Zolnai pointed out that including every possible header resulted in some valid warnings not being reported. Instead, how about just including only our primary include folder of ./include with the '-Iinclude' parameter? This reduced the warnings from 4005 to 523. You can see and compare the results by opening the 'index.html' file at https://drive.google.com/open?id=1HaCVA_udd044uwRNn3RiyGJ01pQdYjwn It seems many valid variableScope warnings are still being omitted. I'm still looking into that. Are there any other categories of valid errors that are missing? Specific examples would be helpful. Overall, does this report have a higher signal-to-noise ratio than our current weekly report? The full command to generate the report was: ../cppcheck/cppcheck -D__GNUC__ -DBOOST_ERROR_CODE_HEADER_ONLY -DBOOST_SYSTEM_NO_DEPRECATED -DCPPU_ENV=gcc3 -DLINUX -DNDEBUG -DOSL_DEBUG_LEVEL=0 -DUNIX -DUNX -DX86_64 -D_PTHREADS -D_REENTRANT -j4 -i external/ -i workdir/ -Iinclude --xml --suppressions-list=cppcheck_supp.txt --enable=all --max-configs=100 ./ 2> ./err.xml ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hi Luke, slacka ezt írta (időpont: 2018. okt. 2., K, 4:05): > The goal of my manual approach was to configure Cppcheck to minimize false > positives. In doing so, I was forced to only scan the Linux code base, as > only Linux has Linux system headers and does not have Window's or BSD's... > So I am not surprised that some valid issues were not reported. > > There are many knobs I could tweak. For example, since my last post, I > discovered I could remove the "-DNDEBUG" to scan the debug code path. I > could also remove the "-j 4" option to allow Cppcheck to scan for unused > functions. I don't know what is most useful, and what valid issues were > not > being reported. This is why I have asked the ML for feedback. > I used to find valid issues amongs the variableScope warnings for example. Check a frequently modified module (e.g. sw, sc, sd). One example: https://dev-builds.libreoffice.org/cppcheck_reports/master/1501.html#line-1729 I just checked a few of these warnings now and they should be there in a Linux specific analysis too. The linked one seems not a platform specific or debug code. So it would be good to find out why your report does not contains this one. In general I doubt that a static analyzer does not find any issue (at least some false positives) in sw module for example (if it was not cleaned up with this analyzer earlier). You report contains no issue in sw. It seems to me the scope of the analysis is greatly reduced by your change, that's why it does not find a lot of issues. I'm not sure how the false positives can be reduced by specifing the includes. Which false positives are coming from wrong includes? Best Regards, Tamás ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
On 02/10/2018 04:05, slacka wrote: There are many knobs I could tweak. For example, since my last post, I discovered I could remove the "-DNDEBUG" to scan the debug code path. I could also remove the "-j 4" option to allow Cppcheck to scan for unused functions. I don't know what is most useful, and what valid issues were not being reported. This is why I have asked the ML for feedback. The knobs you can tweak are the configure options passed to LO's autogen.sh. For example, -DNDEBUG is controlled by --en-/disable-assert-always-abort. There is not per-se an "ideal" set of configure options for static analysis purposes. One reasonable approach is of course to maximize code coverage, both at the large scale (enabling as many optional modules as possible, from e.g. --enable-kde5 to --enable-ext-nlpsolver) and at small (e.g., enabling asserts with --enable-assert-always-abort or debug-only code blocks with --enable-debug or even --enable-dbgutil, both of which imply --enable-assert-always-abort). ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
The goal of my manual approach was to configure Cppcheck to minimize false positives. In doing so, I was forced to only scan the Linux code base, as only Linux has Linux system headers and does not have Window's or BSD's... So I am not surprised that some valid issues were not reported. There are many knobs I could tweak. For example, since my last post, I discovered I could remove the "-DNDEBUG" to scan the debug code path. I could also remove the "-j 4" option to allow Cppcheck to scan for unused functions. I don't know what is most useful, and what valid issues were not being reported. This is why I have asked the ML for feedback. So if a dev wants give me some guidance, I could continue tweaking, or as you suggested, we could run 2 reports. 1) a limited Linux only scan with few false positives (ala my manual approach), and 2) a general scan with many false positives (the current Cppcheck Report). If you try to limit the false positives with include locations without also limiting configuration, Cppcheck gets overloaded and generates tens of thousands of "too many configuration" errors. -- Sent from: http://document-foundation-mail-archive.969070.n3.nabble.com/Dev-f1639786.html ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hi, On Mon, Oct 01, 2018 at 08:52:19AM +0200, Stephan Bergmann wrote: > That smells like gbuild-to-ide needs to be run from within gbuild's > "config_host.mk polluted" environment, which can be done via the "cmd" make > target, i.e., something like > > make cmd cmd='bin/gbuild-to-ide --ide vim --make make' This is exactly what make vim-ide-integration does. :-) (Though this later first runs gbuildtojson.) Regards, Miklos signature.asc Description: Digital signature ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
On 30/09/2018 05:20, Luke Benes wrote: Also when there seems to have been a coding style that all <> includes outside of /inc folders should be defined by their relative path. Cppcheck only complains about 4 missing includes that do not follow this pattern.(see my earlier email on oddball includes). See the root README.md for "Rules for #include directives (C/C++)". ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
On 30/09/2018 15:04, Maarten Hoes wrote: So, my futile attempt was : ./bin/gbuild-to-ide --ide vim --make make Which resulted in this: Traceback (most recent call last): File "./bin/gbuild-to-ide", line 1664, in gbuildparser = GbuildParser(args.makecmd).parse() File "./bin/gbuild-to-ide", line 83, in __init__ self.binpath = os.path.dirname(os.environ['GPERF']) # woha, this is quite a hack File "/usr/lib64/python3.6/os.py", line 669, in __getitem__ raise KeyError(key) from None KeyError: 'GPERF' Appearantly, 'gbuild-to-ide' assumes some environment variables are set (GPERF, SRCDIR, BUILDDIR, INSTDIR, and WORKDIR), but I cannot determine what would be sane/expected values for these variables. Okay, SRCDIR/BUILDDIR/INSTDIR seem self-explanatory (but you know what happens when we assUme), but what are WORKDIR and GPERF ? That smells like gbuild-to-ide needs to be run from within gbuild's "config_host.mk polluted" environment, which can be done via the "cmd" make target, i.e., something like make cmd cmd='bin/gbuild-to-ide --ide vim --make make' ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hi, On Sun, Sep 30, 2018 at 5:20 AM Luke Benes wrote: > Maarten, > Thanks for your suggestion here and your earlier contributions to the > Cppcheck Report. I agree that we should create the include file > dynamically. However the approach used by your script seems like overkill. > Cppcheck already finds that quoted includes like > #include "GraphicExportFilter.hxx" > . > Also when there seems to have been a coding style that all <> includes > outside of /inc folders should be defined by their relative path. Cppcheck > only complains about 4 missing includes that do not follow this > pattern.(see my earlier email on oddball includes). > > Unless, I'm missing something, I still prefer this approach: > $ find . -type d \( -name "inc" -o -name "include" \) |sort > inc.txt > > inc.txt only has ~200 entries, where as /tmp/tmpfile.txt has ~1,800 after > sorting it. > > -Luke > > I was about to speak of the supposedly preferred approach that I took, for the reasons that my approach : 1.) Does not depend on the existence of and/or adherence to any customs/practices/coding styles, which can and do change over time, or are simply overlooked, causing errors that people will have to correct manually or even cause silent failure that no-one notices. 2.) Is fully dynamic, and does not have any paths 'hard-coded' into the script, which one would have to adjust if in the future the paths/names change. I already am worried about the fact that the approach uses hard-coded defines (-DFOO/-UBAR) on the commandline instead of determining them dynamically, but I currently see no way to do this differently, so I guess that has to stay for now. Also, I guess the chance that these defines change over time is far less likely than the chance that directory names and locations change. But then I decided to put my money where my mouth is, and made the necessary changes to the cppcheck-report script, and guess what ? Apart from approach differences, your version completed in an acceptable execution time, whereas my approach took *ages* (after 40 minutes cppcheck was still only about 1% completed, after which I aborted the attempt). As of now, I cannot tell what causes this massive difference in execution time, but the only visible difference in cppcheck output was this : my version: Checking basctl/source/basicide/basides1.cxx: LINUX=1;__GNUC__=1;UNX=1;NDEBUG=1;CORE_LITTLE_ENDIAN=1;__LITTLE_ENDIAN__=1;__x86_64__=1;LINUX=1;__GNUC__=1;UNX=1;NDEBUG=1;CORE_LITTLE_ENDIAN=1;__LITTLE_ENDIAN__=1;__x86_64__=1... your version: Checking basctl/source/basicide/basides1.cxx ... Obvious things: 'my version' listed the -D in the output, your's did not. 'My version' lists the defines *twice* (which can't be good). I don't know what's going on here. - Maarten ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hi, On Sun, Sep 30, 2018 at 11:29 AM Tamás Zolnai wrote: > > I'm not sure what your modification is doing. I just checked the report > you attached and I compared it with the full report and I see your change > filters out not only false positives. > > Am I interpreting correctly here that the proposed approach of doing this : $ find . -type d \( -name "inc" -o -name "include" \) |sort > inc.txt $ cppcheck -DLINUX -D__GNUC__ -DUNX -DNDEBUG -DCORE_LITTLE_ENDIAN -D__LITTLE_ENDIAN__ -D__x86_64__ -UMACOSX -UFREEBSD -U_WIN32 -i external/ -i workdir/ --includes-file=inc.txt --xml --suppressions-list=cppcheck_supp.txt --enable=all --max-configs=100 ./ 2> ./err.xml Versus what the cppcheck-report script currently is doing, results in actual real issues not being reported anymore ? If so, that's bad(TM), and needs further investigation at the very least before that suggested change gets implemented. - Maarten. ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hi, slacka wrote > Ideally, the next step would be to extract the "DEFS": and "INCLUDE": from > gbuild-to-ide and pass that to cppcheck. But that's for another time. Well, the very name 'gbuild-to-ide' sounds intriguing, but I can't figure out what it is supposed to do (and how could it help in the cppcheck-report script context) or how to make it do it's thing (I *really* can't read python code). So, my futile attempt was : ./bin/gbuild-to-ide --ide vim --make make Which resulted in this: Traceback (most recent call last): File "./bin/gbuild-to-ide", line 1664, in gbuildparser = GbuildParser(args.makecmd).parse() File "./bin/gbuild-to-ide", line 83, in __init__ self.binpath = os.path.dirname(os.environ['GPERF']) # woha, this is quite a hack File "/usr/lib64/python3.6/os.py", line 669, in __getitem__ raise KeyError(key) from None KeyError: 'GPERF' Appearantly, 'gbuild-to-ide' assumes some environment variables are set (GPERF, SRCDIR, BUILDDIR, INSTDIR, and WORKDIR), but I cannot determine what would be sane/expected values for these variables. Okay, SRCDIR/BUILDDIR/INSTDIR seem self-explanatory (but you know what happens when we assUme), but what are WORKDIR and GPERF ? - Maarten -- Sent from: http://document-foundation-mail-archive.969070.n3.nabble.com/Dev-f1639786.html ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hello Luke, I'm not sure what your modification is doing. I just checked the report you attached and I compared it with the full report and I see your change filters out not only false positives. I used to use the cppcheck report to give my students a small task for their first patch. So it would be helpful to keep the full report (it can be kept next to the short one I guess), otherwise I would need to run cppcheck myself. Thanks, Tamás Luke Benes ezt írta (időpont: 2018. szept. 30., V, 5:20): > Maarten, > Thanks for your suggestion here and your earlier contributions to the > Cppcheck Report. I agree that we should create the include file > dynamically. However the approach used by your script seems like overkill. > Cppcheck already finds that quoted includes like > #include "GraphicExportFilter.hxx" > . > Also when there seems to have been a coding style that all <> includes > outside of /inc folders should be defined by their relative path. Cppcheck > only complains about 4 missing includes that do not follow this > pattern.(see my earlier email on oddball includes). > > Unless, I'm missing something, I still prefer this approach: > $ find . -type d \( -name "inc" -o -name "include" \) |sort > inc.txt > > inc.txt only has ~200 entries, where as /tmp/tmpfile.txt has ~1,800 after > sorting it. > > -Luke > > ___ > LibreOffice mailing list > LibreOffice@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/libreoffice > ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Maarten, Thanks for your suggestion here and your earlier contributions to the Cppcheck Report. I agree that we should create the include file dynamically. However the approach used by your script seems like overkill. Cppcheck already finds that quoted includes like #include "GraphicExportFilter.hxx" . Also when there seems to have been a coding style that all <> includes outside of /inc folders should be defined by their relative path. Cppcheck only complains about 4 missing includes that do not follow this pattern.(see my earlier email on oddball includes). Unless, I'm missing something, I still prefer this approach: $ find . -type d \( -name "inc" -o -name "include" \) |sort > inc.txt inc.txt only has ~200 entries, where as /tmp/tmpfile.txt has ~1,800 after sorting it. -Luke ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hi, If, by specifying (additional ?) include files/directories and adding defines, you managed to bring down the massive zillions of warnings back to just 30, it seems to me that you managed to do what the script should have been doing all along to begin with. ;-) Just a small remark (and you probably intend to do so already anyway): besides changing the way cppcheck itself is called, you also make use of a 'inc.txt' file (which I assume contains the names of directories that contain include files). I would personally re-create that file dynamically from within the script (instead of providing a static version) on each run, so that if in the future additional similar directories are added (or the directory names change) these will get picked up automatically by the script, instead of someone having to manually re-recreate the 'inc.txt' file by hand (which people might/will forget to do). Yes, this will mean that the script will take slightly longer to complete it's run (my personal measurements based on the example included below are an added 15 seconds or so currently), but since it is run automatically without any live human being waiting for the output as it is running (and it is only run once a week anyway as I recall) this does not really matter. How about something like the thing below ? It would find all header files in the libreoffice source tree, regardless of what directory they are located in, and then use 'dirname' to get just the directory names they are located in, and 'sort -u' to mention each directory just once. Would that be sufficient for your suggested modifications ? - Maarten Hoes #!/bin/sh TMPFILE=/tmp/tmpfile.txt INCLUDEFILE=/tmp/incfile.txt rm -f "$TMPFILE" find . -type f \( -name \*.hxx -o -name \*.h \) | \ while read FILE do DIRNAME=$(dirname "$FILE") echo "$DIRNAME" >> "$TMPFILE" done sort -u "$TMPFILE" > "$INCLUDEFILE" cppcheck --includes-file="$INCLUDEFILE" -- Sent from: http://document-foundation-mail-archive.969070.n3.nabble.com/Dev-f1639786.html ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
As I mentioned before, by manually specifying includes and preprocessor configurations, I was able to reduce the number of warning from ~9000 to 30. https://drive.google.com/file/d/1Ai_Zcj18cdQxVIQESb4lJr5K9LnzjnpW/view?usp=sharing You can view it by unzipping and opening 'index.html'. Caolán or Stephan, Could either one of you take a quick pass at this to see if my improvements are useful? If this is useful, I could modify the weekly Cppcheck Report script to include these improvements. Who could set me up with permissions? -Luke ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Cppcheck: Reduction of False Positives: Manual Approach
Hi! On 9/13/2018 8:59 AM, Luke Benes wrote: By manually specifying includes and preprocessor configurations, I was able to reduce the number of warning from ~9000 to 30. Great! Ideally, the next step would be to extract the "DEFS": and "INCLUDE": from gbuild-to-ide and pass that to cppcheck. But that's for another time. > dbaccess/source/shared/registrationhelper.cxx > 23 preprocessorErrorDirective error #error "don't build this file directly! use dbu_reghelper.cxx instead!" This one is because a .cxx file was processed directly. The workdir/GbuildToJson/Library/dbulo.dll does contain dbaccess/source/shared/dbu_reghelper in its CXXOBJECTS, but no entry for dbaccess/source/shared/registrationhelper. So extracting this information seems helpful, too, to reduce false positives (OTOH I wonder what this inclusion technique is useful here for.) -- Best regards, Mike Kaganski ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Cppcheck: Reduction of False Positives: Manual Approach
By manually specifying includes and preprocessor configurations, I was able to reduce the number of warning from ~9000 to 30. https://drive.google.com/file/d/1Ai_Zcj18cdQxVIQESb4lJr5K9LnzjnpW/view?usp=sharing You can view it by unzipping and opening 'index.html'.Did this uncover any valid issues? The command line used to scan our code base was: $ cppcheck -j4 -DLINUX -D__GNUC__ -DUNX -DNDEBUG -DCORE_LITTLE_ENDIAN -D__LITTLE_ENDIAN__ -D__x86_64__ -UMACOSX -UFREEBSD -U_WIN32 -i external/ -i workdir/ --includes-file=inc.txt --xml --suppressions-list=cppcheck_supp.txt --enable=all --max-configs=100 ./ 2 ./err.xml inc.txt was generated with: $ find . -type d \( -name "inc" -o -name "include" \) |sort inc.txt If this is useful, I could modify the weekly Cppcheck Report script to include these improvements. Who could set me up with permissions? Ideally, the next step would be to extract the "DEFS": and "INCLUDE": from gbuild-to-ide and pass that to cppcheck. But that's for another time. -Luke ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice