Inside a signal handler, you cannot allocate memory because the signal
handler, being implemented with a C
[`signal`](https://en.cppreference.com/w/c/program/signal) call, can be
called _during_ a memory allocation - when that happens, the CTRL-C
handler causes a segfault and/or other inconsistent state.
Similarly, the call can happen from a non-nim thread or inside a C
library function call etc, most of which do not support reentrancy and
therefore cannot be called _from_ a signal handler.
The stack trace facility used in the default handler is unfortunately
beyond fixing without more significant refactoring since it uses
garbage-collected types in its API and implementation.
As an alternative to https://github.com/nim-lang/Nim/pull/25110, this PR
removes the most problematic signal handler, namely the one for SIGINT
(ctrl-c) - SIGINT is special because it's meant to cause a regular
shutdown of the application and crashes during SIGINT handling are both
confusing and, if turned into SIGSEGV, have downstream effects like core
dumps and OS crash reports.
The signal handlers for the various crash scenarios remain as-is - they
may too cause their own crashes but we're already going down in a bad
way, so the harm is more limited - in particular, crashing during a
crash handler corrupts `core`/crash dumps. Users wanting to keep their
core files pristine should continue to use `-d:noSignalHandler` - this
is usually the better option for production applications since they
carry more detail than the Nim stack trace that gets printed.
Finally, the example of a ctrl-c handler performs the same mistake of
calling `echo` which is not well-defined - replace it with an example
that is mostly correct (except maybe for the lack of `volatile` for the
`stop` variable).
Reworked closureiter transformation.
- Convolutedly nested finallies should cause no problems now.
- CurrentException state now follows nim runtime rules (pushes and pops
appropriately), and mimics normal code, which is somewhat buggy, see
#25031
- Previously state optimization (removing empty states or extra jumps)
missed some opportunities, I've reimplemented it to do everything
possible to optimize the states. At this point any extra states or jumps
should be considered a bug.
The resulting codegen (compiled binaries) is also slightly smaller.
**BUT:**
- I had to change C++ reraising logic, see expt.nim. Because with
closure iters `currentException` is not always in sync with C++'s notion
of current exception. From my tests and understanding of C++ runtime
there should not be any problems, but I'm only 99% sure :)
- I've reused `nfNoRewrite` flag in one specific case during the
transformation. This flag is also used in term-rewriting logic. Again,
99% sure, these 2 scenarios will never intersect.
* fix =#13790 ptr char (+friends) should not implicitly convert to cstring
* Apply suggestions from code review
* first round; compiles on windows
* nimPreviewSlimSystem
* conversion is unsafe, cast needed
* fixes more tests
* fixes asyncnet
* another try another error
* last one
* true
* one more
* why bugs didn't show at once
* add `nimPreviewCstringConversion` switch
* typo
* fixes ptr to cstring warnings[backport]
* add fixes
Co-authored-by: xflywind <43030857+xflywind@users.noreply.github.com>
* quit value gets saturated to ranges
* add documentation
* minimal changes
* refactor
* small fix
* add documentation
* fixes
* Update lib/system.nim
Co-authored-by: Juan Carlos <juancarlospaco@gmail.com>
Co-authored-by: Juan Carlos <juancarlospaco@gmail.com>
In CPS we would consume an exception in the except branch by stashing it
into a local then remove the exception from Nim environment so as not to
leak it to other code that would be running before the continuation
continues
However since popCurrentException() assumes that the exception always
exist, removing the exception from an except branch will cause a
SIGSEGV to happen. This commit fixes that.
* IC: renamed to_packed_ast module to ic module
* IC: don't store the --forceBuild flag, makes it easier to test
* IC: enable hello world test
* Codegen: refactorings for IC; changed the name mangling algorithm
* fixed the HCR regressions
* life is too short for HCR
* tconvexhull is now allowed to use deepCopy
* IC exposed a stdlib bug, required a refactoring
* codegen: code cleanups
* IC: even if a module is outdated, its dependencies might come from disk
* IC: progress
* IC: better name mangling, module IDs are not stable
* IC: another refactoring helping with --ic:on --gc:arc
* disable arraymancer on Windows for the time being
* disable arraymancer altogether
* IC: make basic test work with 'nim cpp'
* IC: progress on --ic:on --gc:arc
* wip; name mangling for type info
* nimStackTraceOverride: enable stack traces in exceptions
This is a two-step stack trace collection scheme, because re-raised
exceptions will collect multiple stack traces but use them rarely, when
printing info about an uncaught exception, so it makes sense to only do
the cheap stack unwinding all the time and the relatively expensive
debugging information collection on-demand.
`asyncfutures` implements its own `$` proc for printing
`seq[StackTraceEntry]`, so we have to add the debugging info there, just
like we do for the private `$` proc in `system/excpt`.
* cleaned up PR #15284
Co-authored-by: Ștefan Talpalaru <stefantalpalaru@yahoo.com>
* do not track 'raise Defect' in the .raises: [] clause anymore
* --panics:on maps 'raise Defect' to an unrecoverable fatal error
* make tests green again
* update the documentation too
* Error -> Defect for defects
The distinction between Error and Defect is subjective,
context-dependent and somewhat arbitrary, so when looking at an
exception, it's hard to guess what it is - this happens often when
looking at a `raises` list _without_ opening the corresponding
definition and digging through layers of inheritance.
With the help of a little consistency in naming, it's at least possible
to start disentangling the two error types and the standard lib can set
a good example here.
* stacktraces can now show custom runtime msgs
* improve tests/stdlib/tstackframes.nim
* fix test for --gc:arc
* test --stacktraceMsgs:on and --stacktraceMsgs:off
* --stacktracemsgs:off by default
* config update
* disable a questionable test
* remove c++ exception handling IDs, new impl doesn't require it anymore
* C++ based exceptions finally work
* fixes bootstrapping problem in C++ mode
* teach GCC it's 2020 now
* more bugfixes for C++ based exception handling
* apply cooldome's patch
* another attempt to enable C++11
* bug fix
Co-authored-by: Araq <rumpf_a@web.de>
Co-authored-by: cooldome <ariabushenko@bk.ru>
* os:any implementation
* os:asny: omit flock/funlock calls in echoBinSafe
* Disabled default "unhandled expection" reporting for `--os:any` to reduce
code size. Added unhandledExceptionHook instead which can be used to get
a notification from Nim and handle it from the application.
This implements "deterministic" exception handling for Nim based on goto instead of setjmp. This means raising an exception is much cheaper than in C++'s table based implementations. Supports hard realtime systems. Default for --gc:arc and the C target because it's generally a good idea and arc is all about deterministic behavior.
Note: This implies that fatal runtime traps are not catchable anymore! This needs to be documented.
* libbacktrace support
* switch to a generic stack trace overriding mechanism
When "nimStackTraceOverride" is defined, once of the imported modules
can register its own procedure to replace the default stack trace
generation by calling `registerStackTraceOverride(myOwnProc)`.
Tested with `./koch boot -d:release --debugger:native -d:nimStackTraceOverride --import:libbacktrace`
for the compiler itself and `./bin/nim c -r -f --stacktrace:off --debugger:native -d:nimStackTraceOverride --import:libbacktrace foo.nim`
for an external program.
* make the StackTraceOverrideProc {.noinline.}
* fix undefined behavior in terminate handler
* fix failing unit test
* Revert "fix failing unit test"
This reverts commit 0e5e385fbf.
* Revert "fix undefined behavior in terminate handler"
This reverts commit 2b582871f1.
* do not throw inside terminate handler with msvc < 1923
