Trying to avoid that round-trip on versions that are known to
not have it puts us on the mercy of detecting the version,
which is unreliable by nature, and bad in practice, as proven
by the recent MinGW builds.
Change-Id: I122eb03fe0ddcb193ed8d45e672564f7113b662f
Reviewed-by: Ray Donnelly <mingw.android@gmail.com>
Reviewed-by: hjk <hjk121@nokiamail.com>
We need to distiguish between start of GDB itself (a.k.a .gdbinit-style
customization) and after the dumper machinery is initialized, which is
nowadays often delayed until the first stop hook.
Change-Id: I40f1e7225c2043b8bcb7d50eef948bb3c9162bb6
Reviewed-by: Leena Miettinen <riitta-leena.miettinen@digia.com>
Reviewed-by: hjk <hjk121@nokiamail.com>
Version of gdb < 7.4 means we are on Mac, so we can
disable commands that have known results there.
Change-Id: I8fc97ed61c47af2c3d9e5cc2bf83e97661204d4f
Reviewed-by: hjk <hjk121@nokiamail.com>
We don't support GDB in the range 7.0...7.3 anymore where the
outcome was unsure.
Change-Id: I1ce76f5c4b1febe45c4a3b6404208fed7d9c6868
Reviewed-by: hjk <hjk121@nokiamail.com>
See sourceware.org/bugzilla/show_bug.cgi?id=12023.
The workaround can cause hickups in the output parser
of Mac gdb, as the error message is scattered within
regular output without any markup.
Change-Id: I835133e390403a0f8f6aa573d6f9c63cf0784187
Reviewed-by: hjk <hjk121@nokiamail.com>
Reduce compatibility issues e.g. on Mac. The previous versions
are not working well anyway.
Change-Id: If6cc7c1d4a85300d9bf054966216b52b35c0d1f2
Reviewed-by: hjk <hjk121@nokiamail.com>
This essentially reverts 3aa0e50039 , which causes trouble
e.g. for network paths and other corner cases.
Change-Id: I12ceb5fbc68b7a3fc8010165acc910acd2d95e55
Reviewed-by: hjk <hjk121@nokiamail.com>
Breakpoints by path are not hit if the path traverses
symbolic links.
Change-Id: I3ab5d1bead8333986444aec709c940840dfddbb4
Reviewed-by: hjk <hjk121@nokiamail.com>
Setting breakpoints by file name and line does not work robustly when
using gdb, with neither the full path, nor with just the baseline. Since
typically one of the options work, Creator offers a per-breakpoint choice
of which approach to use. Traditionally, using the full path broke more
often, so the default was "short". Lately (gdb 7.3+) using the full path
seems to be more robust, so the default should be changed. Manual
overriding to "short" is still possible.
Change-Id: I9e857c86a63964bdacf9bebc5444ea752e5974f8
Reviewed-by: hjk <hjk121@nokiamail.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
- Split off GDB and LLDB specific parts into separate files,
- Trigger loading of dumpers.py and qttypes.py from bridge.py
- Read start up script at startup, not as part of dumper loading
Change-Id: I7941ee535121fa0f43a466e5bb75a18c9bb19764
Reviewed-by: hjk <hjk121@nokiamail.com>
This enables the use "official gdb"-styly pretty printers from
within Creator. In case there are both variants, Creator-style
pretty printers are preferred.
Change-Id: Iaa178d922dae4eaa1943e9d5b960a666f80fe4db
Reviewed-by: hjk <hjk121@nokiamail.com>
This only worked for distributions with a RedHat/Fedora upstream
so far and requires more dynamicity in the TaskWindow as the
TaskHandler dies with the engine, leading to crashs.
Change-Id: I20d0b3c79e4fd0026cd2a0c9ff5b0a4be22c3121
Reviewed-by: Tobias Hunger <tobias.hunger@digia.com>
As we disable of 'by-address' breakpoints by default after each
debugger run as they are unlikely to be correct on the next run,
starting with disabled breakpoints might happen. Disabling it
immediately is faster and less racy than relying on the 'change
needed' mechanism.
Change-Id: Icc5548d322f7ef4f099d218b4f04b4c40e3cb850
Reviewed-by: David Schulz <david.schulz@digia.com>
This avoids a problem when setting the ignore count on a pending
breakpoint.
Task-number: QTCREATORBUG-8809
Change-Id: I8cdeb9c525de33336a21c739c6dc99f824f6fe7a
Reviewed-by: David Schulz <david.schulz@digia.com>
Modern versions of gdb support this and it avoids a nasty
endless loop when setting a condition on a breakpoint in
a non-existing file.
Change-Id: Id3a03a44c016f85617bb1e74a4843b1b398c490f
Reviewed-by: Nikolai Kosjar <nikolai.kosjar@digia.com>
This is a recent regression introduced by the delayed python dumper
initialization.
Task-number: QTCREATORBUG-8615
Change-Id: I3b5dbe5b04c68cf873e3a2597edc60fe2888b3c9
Reviewed-by: Orgad Shaneh <orgads@gmail.com>
This is now in sourceutils.{h,cpp} to make watchutils.{h,cpp}
better acessible to the debugger auto-tests.
Change-Id: Ie87e715bc7018ca190a460c37dfd19bc897059f0
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
This skips the case of "disassembling a function given an address"
with indeterministic time behaviour and and immediately falls back
to the fixed range version instead.
Change-Id: I90ec56198a7d767bcf0a9ffb030637c2ee5c617a
Reviewed-by: Kai Koehne <kai.koehne@digia.com>
