D issues are now
tracked on GitHub. This Bugzilla instance remains as a read-only archive.
Summary: |
while(true) loop with try/catch block causes segfault |
Product: |
D
|
Reporter: |
Brian Palmer <brian+d> |
Component: |
dmd | Assignee: |
No Owner <nobody> |
Status: |
RESOLVED
FIXED
|
Severity: |
critical
|
CC: |
bugzilla, clugdbug, hsteoh
|
Priority: |
P2
|
Version: |
D2 |
Hardware: |
All |
OS: |
Mac OS X |
I ran into a segfault when using a try/catch block inside a while(true) loop when there are other statements after the try/catch block. I reduced it down to this consistent segfault:
import std.stdio;
void main() {
while (true) {
try {
throw new Exception("ah");
} catch (Exception e) {
}
writeln("remove this line and it won't segfault");
}
}
This segfaults in the first iteration on DMD 2.058 on OS X 10.7.3. It segfaults with both -m32 and -m64. Stacktrace:
0 libdyld.dylib 0x00007fff8f3436cd misaligned_stack_error_entering_dyld_stub_binder + 0
1 ??? 0x000000010f535030 0 + 4552085552
2 bug1.d.DC445724916B27653F5E50703AE848DF 0x000000010f51f133 _Dmain + 103
3 bug1.d.DC445724916B27653F5E50703AE848DF 0x000000010f52f8dd D2rt6dmain24mainUiPPaZi7runMainMFZv + 29
4 bug1.d.DC445724916B27653F5E50703AE848DF 0x000000010f52f28d D2rt6dmain24mainUiPPaZi7tryExecMFMDFZvZv + 45
5 bug1.d.DC445724916B27653F5E50703AE848DF 0x000000010f52f92f D2rt6dmain24mainUiPPaZi6runAllMFZv + 63
6 bug1.d.DC445724916B27653F5E50703AE848DF 0x000000010f52f28d D2rt6dmain24mainUiPPaZi7tryExecMFMDFZvZv + 45
7 bug1.d.DC445724916B27653F5E50703AE848DF 0x000000010f52f210 main + 232
8 bug1.d.DC445724916B27653F5E50703AE848DF 0x000000010f51f0c4 start + 52
Removing the writeln, or changing it to a non-infinite loop like int x;while(++x<3) will avoid the segfault, so it seems likely DMD is making a bad assumption about the loop related to it being infinite.
I can't seem to reproduce this problem with the latest git dmd (2.059 candidate) on Linux AMD64. Could it be a Mac-specific problem?
(In reply to comment #1)
> I can't seem to reproduce this problem with the latest git dmd (2.059
> candidate) on Linux AMD64. Could it be a Mac-specific problem?
Looks like it, based on the first location in the backtrace:
misaligned_stack_error_entering_dyld_stub_binder
The stack isn't getting aligned properly before the call to writeln.
Probably related to bug 1115.