Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
PaddlePaddle
Paddle
提交
2469b578
P
Paddle
项目概览
PaddlePaddle
/
Paddle
接近 2 年 前同步成功
通知
2323
Star
20933
Fork
5424
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1423
列表
看板
标记
里程碑
合并请求
543
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
1,423
Issue
1,423
列表
看板
标记
里程碑
合并请求
543
合并请求
543
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
未验证
提交
2469b578
编写于
7月 29, 2020
作者:
C
Chen Weihang
提交者:
GitHub
7月 29, 2020
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Unified paddle error format when catch system signal (#25765)
* unified signal error format * refine signal error message
上级
818d38f1
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
57 addition
and
16 deletion
+57
-16
paddle/fluid/platform/init.cc
paddle/fluid/platform/init.cc
+57
-16
未找到文件。
paddle/fluid/platform/init.cc
浏览文件 @
2469b578
...
@@ -229,25 +229,66 @@ void InitDevices(bool init_p2p, const std::vector<int> devices) {
...
@@ -229,25 +229,66 @@ void InitDevices(bool init_p2p, const std::vector<int> devices) {
}
}
#ifndef _WIN32
#ifndef _WIN32
// Description Quoted from
// https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/signal.h.html
const
struct
{
const
char
*
name
;
const
char
*
error_string
;
}
SignalErrorStrings
[]
=
{
{
"SIGSEGV"
,
"Segmentation fault"
},
{
"SIGILL"
,
"Illegal instruction"
},
{
"SIGFPE"
,
"Erroneous arithmetic operation"
},
{
"SIGABRT"
,
"Process abort signal"
},
{
"SIGBUS"
,
"Access to an undefined portion of a memory object"
},
{
"SIGTERM"
,
"Termination signal"
},
};
bool
StartsWith
(
const
char
*
str
,
const
char
*
prefix
)
{
size_t
len_prefix
=
strlen
(
prefix
);
size_t
len_str
=
strlen
(
str
);
return
len_str
<
len_prefix
?
false
:
memcmp
(
prefix
,
str
,
len_prefix
)
==
0
;
}
const
char
*
ParseSignalErrorString
(
const
std
::
string
&
str
)
{
for
(
size_t
i
=
0
;
i
<
(
sizeof
(
SignalErrorStrings
)
/
sizeof
(
*
(
SignalErrorStrings
)));
++
i
)
{
if
(
std
::
string
::
npos
!=
str
.
find
(
SignalErrorStrings
[
i
].
name
))
{
return
SignalErrorStrings
[
i
].
error_string
;
}
}
return
"Unknown signal"
;
}
// Handle SIGSEGV, SIGILL, SIGFPE, SIGABRT, SIGBUS, and SIGTERM.
std
::
ostringstream
signal_msg_dumper
;
void
SignalHandle
(
const
char
*
data
,
int
size
)
{
void
SignalHandle
(
const
char
*
data
,
int
size
)
{
auto
file_path
=
string
::
Sprintf
(
"/tmp/paddle.%d.dump_info"
,
::
getpid
());
try
{
try
{
// The signal is coming line by line but we print general guide just once
// NOTE1: The glog FailureSignalHandler dumped messages
std
::
call_once
(
glog_warning_once_flag
,
[
&
]()
{
// are deal with line by line
LOG
(
WARNING
)
<<
"Warning: PaddlePaddle catches a failure signal, it may "
// NOTE2: we only deal with the time info ane signal info,
"not work properly
\n
"
;
// the stack trace will generated by paddle self
LOG
(
WARNING
)
<<
"You could check whether you killed PaddlePaddle "
if
(
StartsWith
(
data
,
"*** Aborted at"
))
{
"thread/process accidentally or report the case to "
signal_msg_dumper
<<
" [TimeInfo: "
<<
std
::
string
(
data
,
size
-
1
)
"PaddlePaddle
\n
"
;
<<
"]
\n
"
;
LOG
(
WARNING
)
<<
"The detail failure signal is:
\n\n
"
;
}
else
if
(
StartsWith
(
data
,
"***"
))
{
});
std
::
string
signal_info
(
data
,
size
-
1
);
std
::
string
useless_substr
(
"; stack trace:"
);
LOG
(
WARNING
)
<<
std
::
string
(
data
,
size
);
size_t
start_pos
=
signal_info
.
rfind
(
useless_substr
);
std
::
ofstream
dump_info
;
signal_info
.
replace
(
start_pos
,
useless_substr
.
length
(),
""
);
dump_info
.
open
(
file_path
,
std
::
ios
::
app
);
signal_msg_dumper
<<
" [SignalInfo: "
<<
signal_info
<<
"]
\n
"
;
dump_info
<<
std
::
string
(
data
,
size
);
// NOTE3: Here does not throw an exception,
dump_info
.
close
();
// otherwise it will casue "terminate called recursively"
auto
exp
=
platform
::
EnforceNotMet
(
platform
::
errors
::
Fatal
(
"A serious error (%s) is detected by the operating system."
,
ParseSignalErrorString
(
signal_info
)),
__FILE__
,
__LINE__
);
std
::
cout
<<
exp
.
what
()
<<
signal_msg_dumper
.
str
()
<<
std
::
endl
;
}
}
catch
(...)
{
}
catch
(...)
{
// Since the program has already triggered a system error,
// no further processing is required here, glog FailureSignalHandler
// will Kill program by the default signal handler
}
}
}
}
#endif
#endif
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录