[support]

Hi,

Please try this build: http://visualgdb.com/tmp/VisualGDB-5.0.3.312.msi

If cleaning is still not working, please provide a build log and double-check that the gradlew.bat file is present in the directory that VisualGDB reports.

If android:debuggable is still added, please check the build log, it should now print the full path of the AndroidManifest.xml it is editing. Please let us know which file version is VisualGDB trying to edit.

We used zipalign because assemble did not seem to work properly on some older version of Gradle. As it seems to work now, we have changed it to assemble and added a setting that allows overriding it.

[support]

Please double-check the COM port settings in the Windows Device Manager and ensure that the speed and the parity settings match the settings on the Linux side.

[support]

The general outline of building a cross-toolchain is described in http://gnutoolchains.com/building.

However in most of the cases you will face multiple build errors, as the GNU tools are quite fragile and often incompatible and require ad hoc fixing. We do offer a custom toolchain building service. I.e. if you provide us with a snapshot of your Linux system, we would build an equivalent Windows cross-toolchain for you. Please contact our sales to get a quote for that.

You can alternatively use VisualGDB with clang. As clang command-line options are compatible with gcc, simply replace the gcc executables with clang executables in debug.mak and release.mak

[support]

Hi,

Please try the following build: http://visualgdb.com/tmp/VisualGDB-5.0.3.310.msi

We have made the following changes:

• When using Gradle build subsystem, VisualGDB now uses the $(SourceDir)\$(GradleAppSubdirectory)\build\intermediates\manifests\full\$(GradleFlavor)\$(GradleBuildName)\AndroidManifest.xml file to determine the package/activity names and other settings. I.e. all changes caused by the Gradle flavor will be now read correctly. The path (that appears to be hardcoded in Gradle) can be currently modified by editing the .vgdbsettings file manually.
• VisualGDB now uses the zipalign$(GradleFlavor)$(GradleBuildName) as the gradle task name. The GradleFlavor can be defined via VisualGDB Project Properties as before.
• VisualGDB now does not edit the AndroidManifest.xml file if it has successfully found a generated manifest file from the intermediates folder.
• You can specify additional arguments to Gradle by adding a VisualGDBProjectSettings2/Build/AdditionalGradleArguments element to the .vgdbsettings file (no GUI setting for it yet).
• In case of multiple apk files VisualGDB checks whether the $(app subdirectory)-$(gradle flavor)-Debug.apk file is present. If it is found, it is automatically selected without showing any extra windows.
• If VisualGDB has to rebuild the .jar file, it now displays the file that caused the rebuild.

The build output is actually still available if you select ‘build’ in the “show output from” field. Opening VisualGDB Project Properties switches the output view to VisualGDB internal output.

[support]

Looks like your kernel is configured without kgdb support. Please enable it via kernel configuration interface and then build kgdboe again with the new headers.

[support]

The SIGSEGV is most likely caused by the workaround VisualGDB utilizes to fix issues with library symbol loading. Please try switching the workaround method or disabling it completely.

VisualGDB rebuilds the .jar file when any of the .so files that go into it has a more recent timestamp than the .jar file itself. Please double-check why your .so files get rebuilt.

Regarding the gradle flavor issues, we will do a full investigation/analysis of that after we release the new clang-based IntelliSense engine. Before that we should be able to provide hotfixes that should make debugging possible with reasonable manual adjustments. Could you make and send us a small small repro case demostrating the broken build/launch for a custom flavor? We should be able to release a hotfix based on it very quickly.

[support]

Hi,

As your scenario (debugging a completely custom kernel) is very rare, we will most likely not create a separate product supporting it. However what we can do is extend the VisualGDB plugin interface so that you could add your custom tool windows and use the convenient VisualGDB SDK interfaces to query various data from the GDB session. Let us know if that works for you.

[support]

Looks like the LinuxProject1.cpp file did not get transferred to the remote machine and hence GNU Make cannot find it. Please double-check your file transfer settings and ensure this file exists on the Windows side.

[support]

Hi,

The seat ID is shown directly in the form that displays the link to the activation page.

[support]

Hi,

Please try the newest VisualGDB 5.0 Preview 3 build. It supports customizing the Gradle flavor via VisualGDB Project Properties dialog.

[support]

You cannot reliably combine code from MS C++ compiler and GCC in one module as those compilers have too many differences. However you can create shared library (.DLL) using VisualGDB and load it from your MSVC-compiled app (or vice versa). As long as those are 2 different modules, it should work.

[support]

Hi,

Simply add “-j <number of cores>” to the Make command line arguments in VisualGDB Project Properties to enable parallel building.

[support]

The easiest solution would be to pass the directory of the main makefile and then add further files to your project via <right click in Solution Explorer>->add->Add folder recursively.

[support]

Yes, you can follow this tutorial: http://visualgdb.com/tutorials/raspberry/qt-cross/

[support]

Hi,

Sorry for the delayed reply. Please find the answers to your questions below:

1. If you are using VisualGDB 4.3r4, the TOOLCHAIN_ROOT should be set on-the-fly based on the current project toolchain. Please double-check that it’s defined with the ?= syntax in mcu.mak and that the environment variable is passed correctly (add ‘set’ to build commands in Makefile to get a dump of the environment).
2. Visual Studio project variable system is different from VisualGDB variable system, so the IntelliSense includes cannot reference VisualGDB variables. Simply specify the relative paths to your include directories (e.g. ..\include) and they will be handled correctly.
3. This is a known issue that has been fixed in the upcoming 5.0 release.
4. This is caused by a bug in the texane-STLink tool that VisualGDB uses. Simply unplug your ST-Link and plug it back to get it fixed or switch to OpenOCD that does not have this problem.
5. This is a limitation of VS. VisualGDB projects are Makefile projects, so Visual Studio has no way of determining whether any of the files have changed. We do have long-term plans of transparently supporting MSBuild, that will eliminate this problem, but currently the only workaround is to answer ‘no’ if you are 100% sure that the project has not been modified. VisualGDB will then debug the previously built binary.
6. There is no special procedure. The installer simply unpacks an archive and registers the toolchain with VisualGDB. Simply delete the toolchain folder and VisualGDB will recognize that it’s gone.
7. We do have plans to support HAL for other device families and also allow selecting between HAL and classic frameworks, however this is planned after the release of the new IntelliSense engine.
8. You can easily achieve that by adding ‘-jXXX’ to VisualGDB Project Properties -> Build Settings -> Additional Arguments.
9. This might be quite tricky to support properly (GDB/VS won’t support it natively, so we would have to add some special GUI for that) and it will only work with GDB stubs that provide a command interface that is separate from GDB. You should be able to achieve a similar result if you use OpenOCD, telnet to port 4444 while it is running, and run commands like “mdw <addr> <count>” to query memory contents directly from it (bypassing VS and GDB). Let us know if that works for you.
10. We did experiment with ST-Link/V2 and SWW (built an open-source trace tool from https://github.com/obe1line/stlink-trace) and found out that at seemingly random points in time the ST-Link device enters some internal state where it simply starts producing garbage instead of the trace output. Furthermore, different firmware versions showed different behavior, so we gave up. If you are willing to experiment with it and could get the tool to work reliably, we could help you port its functionality to OpenOCD so that VisualGDB would support it directly. You can find the source with our modifications to get it working on STM32F4Discovery here: http://visualgdb.com/tmp/trace.zip (Disclaimer: we are not related to the author of the tool and find parts of its code as puzzling as you will).

11. (Regarding the <MCUDefinitionFile>). Thanks for letting us know, we will include a fix in the upcoming 5.0 release.

[Author]

Posts