Skip to main content

Notice: this Wiki will be going read only early in 2024 and edits will no longer be possible. Please see: https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/wikis/Wiki-shutdown-plan for the plan.

Jump to: navigation, search

Difference between revisions of "TCF/DraftPlan12"

< TCF
Line 15: Line 15:
  
 
==Proposed Themes==
 
==Proposed Themes==
 +
 +
===Robustness and Matrix Coverage===
 +
''Interested Parties:'' All
 +
 +
* We will continue making TCF a rock-solid commercial-grade debug solution and increase its coverage of target architectures and compilers.
 +
** Increase coverage of supported ELF tags and ELF format variants in the symbol and DWARF readers
 +
** Fix defects and increase test coverage
  
 
===Improve Unittest coverage at Eclipse.org===
 
===Improve Unittest coverage at Eclipse.org===
 
''Interested Parties:'' Wind River, Xilinx
 
''Interested Parties:'' Wind River, Xilinx
  
* We have existing unit tests of the Java API, but agent features are currently not routinely tested in Open Source
+
* We have existing unit tests of the Java API running on Eclipse.org, and the agent code is routinely tested at various TCF contributors.
** We'll want to set up an infrastructure for automatically building the TCF C agent, and having it unit-tested
+
** We'll want to set up an infrastructure for automatically building the TCF C agent, and having it unit-tested on Eclipse.org servers for more transparency of the process.
 
** We'll strive for an automated integration test of the Java vs. C bits on Eclipse.org servers.
 
** We'll strive for an automated integration test of the Java vs. C bits on Eclipse.org servers.
 +
** We'll strive for an automated integration test of the Python API's.
  
 
===Remote-controlling Eclipse via TCF===
 
===Remote-controlling Eclipse via TCF===

Revision as of 20:49, 9 September 2013

TCF 1.2 DRAFT plan

Compatibility

We will be shooting for a TCF 1.2 version, ie keep API's compatible:

  • TCF protocol must remain 100% compatible with TCF 1.1 (it's part of the TCF value proposition)
  • TCF Java API must remain 100% binary compatible (we have clients coding against it)
  • TCF Python and C API should remain 100% compatible (we have clients but this is source compatibility so they can rebuild or use version checks)

Timing

  • TCF 1.2 shall release with Eclipse Luna in June 2014
  • We may want a TCF release earlier, then TCF 1.2 releases earlier and TCF 1.3 comes with Eclipse Luna

Proposed Themes

Robustness and Matrix Coverage

Interested Parties: All

  • We will continue making TCF a rock-solid commercial-grade debug solution and increase its coverage of target architectures and compilers.
    • Increase coverage of supported ELF tags and ELF format variants in the symbol and DWARF readers
    • Fix defects and increase test coverage

Improve Unittest coverage at Eclipse.org

Interested Parties: Wind River, Xilinx

  • We have existing unit tests of the Java API running on Eclipse.org, and the agent code is routinely tested at various TCF contributors.
    • We'll want to set up an infrastructure for automatically building the TCF C agent, and having it unit-tested on Eclipse.org servers for more transparency of the process.
    • We'll strive for an automated integration test of the Java vs. C bits on Eclipse.org servers.
    • We'll strive for an automated integration test of the Python API's.

Remote-controlling Eclipse via TCF

Interested Parties: Wind River

  • TCF has been designed for remote-controlling embedded systems from a host-side tool. But for many end-to-end scripting scenarios, it is also interesting to control the host-side tool from a target or from an external scripting commandline.
    • We will contribute a TCF Scripting API that allows any peer to expose functionality to be scripted through a TCF communication channel, along with the ability to discover scriptable commands, help texts, and code completion.
    • As exemplary implementation, we will provide an Eclipse scripting server that exposes Eclipse functionality registered through extension point for scripting from the outside
    • As exemplary client, we will provide a simple "plain Java" commandline application that allows scripting Eclipse from the outside
    • Initial exemplary scripted actions will include show-view, and open-perspective.


Finish ARM Debug Support for Linux ptrace

Interested Parties: Wind River

  • ARM is an important embedded architecture, TCF should provide good debug support
    • Single-stepping will need some code that involves parsing opcodes

Collaborate with Linux Distro Builders

Interested Parties: Wind River

  • For widespread adoption, a TCF agent should come out-of-the-box with Yocto, Buildroot, Raspbian and potentially other Linux distros.

Collaborate with gdb and Multicore Association

Interested Parties: Wind River, Xilinx

  • gdb is very strong thanks to its rich and well-known command-line and scripting support; but it has performance problems on multicore embedded systems since the "gdbserver" protocol is very limited. The gdb community is thinking about a new remote protocol. TCF seems to exactly match their needs. Collaboration should be considered.
  • The Multicore Association TIWG is working on a generic protocol for tooling. TCF seems to be a very good match for their needs. Continue exchange and collaboration.

Further Improve Usability

Interested Parties: Wind River

  • We'll want to further improve the out-of-box experience and usability for TCF-based products.
  • Add more HOWTO's and out-of-box documentation instructions like the Raspberry Pi HOWTO
  • Consider a pre-built package similar to the Buildroot Eclipse Tools Package
  • TCF is strong in language bindings and scripting. Provide more examples for how to use the TCF Python and Lua bindings.

Tracing and Profiling API's

Interested Parties: Freescale, Xilinx, WindRiver

  • There is interest from multiple parties in adding standardized tracing and profiling API's to TCF.

Add PowerPC Debug Support for Linux ptrace

Interested Parties:

  • Having PPC debug support in addtion to Intel and ARM would likely further help adoption with embedded distros.

Back to the top