[Buildroot] [PATCH] package/google-breakpad: fix build on modern hosts

* [Buildroot] [PATCH] package/google-breakpad: fix build on modern hosts
@ 2019-12-26 21:45 Thomas Petazzoni
  2019-12-30 21:13 ` Yann E. MORIN
  0 siblings, 1 reply; 2+ messages in thread
From: Thomas Petazzoni @ 2019-12-26 21:45 UTC (permalink / raw)
  To: buildroot

Since glibc 2.30, a tgkill() function is exposed by the C library, but
google-breakpad has its own internal definition of it, which now
conflicts. This causes build failures on modern build machines (when
building google-breakpad for the host).

This commit adds a patch that simply renames the internal tgkill()
function to BreakpadTgkill() to avoid the naming conflict.

We do that instead of a configure.ac change to avoid having to
autoreconf this package, and because the fix is anyway not
upstreamable as upstream simply dropped the internal tgkill()
implementation, but using that would break building google-breakpad on
older systems.

Fixes:

  http://autobuild.buildroot.net/results/bc2ae827b830d23094c8b70e5b34911d060295a3/ (host)
  http://autobuild.buildroot.net/results/21257e5a87f41487c6bf4db4e15ce49f1af1ac1e/ (target)

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
---
 ...-handler-exception_handler.cc-rename.patch | 59 +++++++++++++++++++
 1 file changed, 59 insertions(+)
 create mode 100644 package/google-breakpad/0003-src-client-linux-handler-exception_handler.cc-rename.patch

diff --git a/package/google-breakpad/0003-src-client-linux-handler-exception_handler.cc-rename.patch b/package/google-breakpad/0003-src-client-linux-handler-exception_handler.cc-rename.patch
new file mode 100644
index 0000000000..579e7f3457
--- /dev/null
+++ b/package/google-breakpad/0003-src-client-linux-handler-exception_handler.cc-rename.patch
@@ -0,0 +1,59 @@
+From 2fa414c8655c421e7eb0bb1719928babb0ecf7c6 Mon Sep 17 00:00:00 2001
+From: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
+Date: Thu, 26 Dec 2019 22:21:33 +0100
+Subject: [PATCH] src/client/linux/handler/exception_handler.cc: rename tgkill
+ to BreakpadTgkill()
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+Since glibc 2.30, a tgkill() function was added in the C library, and
+its definition obviously conflicts with the internal definition of
+google-breakpad, causing build failures:
+
+src/client/linux/handler/exception_handler.cc:109:12: error: ?int tgkill(pid_t, pid_t, int)? was declared ?extern? and later ?static? [-fpermissive]
+  109 | static int tgkill(pid_t tgid, pid_t tid, int sig) {
+      |            ^~~~~~
+In file included from /usr/include/signal.h:374,
+                 from ./src/client/linux/handler/exception_handler.h:33,
+                 from src/client/linux/handler/exception_handler.cc:66:
+/usr/include/bits/signal_ext.h:29:12: note: previous declaration of ?int tgkill(__pid_t, __pid_t, int)?
+   29 | extern int tgkill (__pid_t __tgid, __pid_t __tid, int __signal);
+      |            ^~~~~~
+
+Upstream google-breakpad simply dropped the use of the internal
+tgkill() in commit
+https://chromium.googlesource.com/breakpad/breakpad/+/7e3c165000d44fa153a3270870ed500bc8bbb461. However,
+this is not realistic for Buildroot, since we do support old systems
+where the system C library will not necessarily provide tgkill().
+
+Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
+---
+ src/client/linux/handler/exception_handler.cc | 4 ++--
+ 1 file changed, 2 insertions(+), 2 deletions(-)
+
+diff --git a/src/client/linux/handler/exception_handler.cc b/src/client/linux/handler/exception_handler.cc
+index b63f973b..b4c279b8 100644
+--- a/src/client/linux/handler/exception_handler.cc
++++ b/src/client/linux/handler/exception_handler.cc
+@@ -106,7 +106,7 @@
+ #endif
+ 
+ // A wrapper for the tgkill syscall: send a signal to a specific thread.
+-static int tgkill(pid_t tgid, pid_t tid, int sig) {
++static int BreakpadTgkill(pid_t tgid, pid_t tid, int sig) {
+   return syscall(__NR_tgkill, tgid, tid, sig);
+   return 0;
+ }
+@@ -387,7 +387,7 @@ void ExceptionHandler::SignalHandler(int sig, siginfo_t* info, void* uc) {
+     // In order to retrigger it, we have to queue a new signal by calling
+     // kill() ourselves.  The special case (si_pid == 0 && sig == SIGABRT) is
+     // due to the kernel sending a SIGABRT from a user request via SysRQ.
+-    if (tgkill(getpid(), syscall(__NR_gettid), sig) < 0) {
++    if (BreakpadTgkill(getpid(), syscall(__NR_gettid), sig) < 0) {
+       // If we failed to kill ourselves (e.g. because a sandbox disallows us
+       // to do so), we instead resort to terminating our process. This will
+       // result in an incorrect exit code.
+-- 
+2.24.1
+
-- 
2.24.1

^ permalink raw reply related	[flat|nested] 2+ messages in thread