[LLVMbugs] [Bug 15802] New: On windows %INCLUDE% ending in \ leads to clang hanging when trying to kick of new process
bugzilla-daemon at llvm.org
bugzilla-daemon at llvm.org
Sat Apr 20 16:48:05 PDT 2013
http://llvm.org/bugs/show_bug.cgi?id=15802
Bug ID: 15802
Summary: On windows %INCLUDE% ending in \ leads to clang
hanging when trying to kick of new process
Product: clang
Version: trunk
Hardware: PC
OS: Windows NT
Status: NEW
Severity: normal
Priority: P
Component: -New Bugs
Assignee: unassignedclangbugs at nondot.org
Reporter: van_kessel at freenet.de
CC: llvmbugs at cs.uiuc.edu
Classification: Unclassified
Created attachment 10389
--> http://llvm.org/bugs/attachment.cgi?id=10389&action=edit
possible patch
On Windows, invoking "clang test.c" makes clang hang during startup if
%INCLUDE% ends in a \ and presumably some similar conditions as well.
The problem has been described by user qble on
http://stackoverflow.com/questions/15992645/clang-compiler-hangs-on-windows
When clang gets invoked with actual work, it assembles a list of arguments and
kicks of a new process. Among those arguments is the include environment
variable on windows. In case this variable contains spaces or some other (e.g.
"C:\Program Files\SomeVendor\Include\") or some other special characters, clang
adds quotation marks around this string so that it stays a single argument to
the new process. If the argument ends in an uneven number (typically one) of
backslashes though, the quotation mark becomes escaped and thus instead of a
invocation like
clang "C:\Program Files\SomeVendor\Include\\" someOtherArg
which makes
argv[1] = "C:\\Program Files\\SomeVendor\\Include\\"
argv[2] = "someOtherArg"
it becomes
clang "C:\Program Files\SomeVendor\Include\" someOtherArg
which is a unterminated
argv[1] = "C:\\Program Files\\SomeVendor\\Include\" someOtherArg ...
The problem can easily be fixed by escaping uneven trailing backslashes by just
adding another one. I have added a pach to llvm\lib\Support\Windows which fixes
the problem, I didn't feel like messing around with c-strings like the rest of
that code does. If you feel strongly about it, I could also write a patch in
keeping with the c-style code and without unnecessary copies (though I don't
think that code is remotely performance critical)
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20130420/63741a8e/attachment.html>
More information about the llvm-bugs
mailing list