| Summary: | Regression: hook-continue no longer gets called by exec-continue | ||
|---|---|---|---|
| Product: | gdb | Reporter: | Andreas Fritiofson <andreas.fritiofson> |
| Component: | mi | Assignee: | Not yet assigned to anyone <unassigned> |
| Status: | UNCONFIRMED --- | ||
| Severity: | normal | CC: | gdb-prs |
| Priority: | P2 | ||
| Version: | 6.8 | ||
| Target Milestone: | 6.8 | ||
| Host: | i686-mingw32 | Target: | arm-none-eabi |
| Build: | Last reconfirmed: | ||
I have a few hooks in my .gdbinit to take care of target specific tweaks when stopping/resuming. In particular, interrupts should be disabled when stepping, but not when running: define hook-stop mon cortex_m3 maskisr on end define hook-continue mon cortex_m3 maskisr off end define hook-until mon cortex_m3 maskisr off end This worked great in gdb 6.7, even when gdb was controlled via the MI interface (from eclipse). In 6.8 however, the continue hook has stopped working over the MI interface. It still works from the CLI and other hooks, such as the above hook-until, works even over MI ("run-to-line" correctly unmasks interupts in eclipse, but "resume" doesn't). I suspect the culprit is changing mi_cmd_exec_continue() (mi/mi-main.c) from being a wrapper around "continue" to using libgdb. ( http://sourceware.org/cgi-bin/cvsweb.cgi/src/gdb/mi/mi-main.c.diff?r1=1.125&r2=1.126&cvsroot=src ) Other commands such as exec-next are still wrappers and their hooks work, but they have FIXME notices to switch to libgdb, so I fear they won't in the future.