xref: /petsc/src/sys/error/errtrace.c (revision 10450e9e44b354a0a3da7bbd573407bdf051df10)
10039db0dSBarry Smith #define PETSC_DESIRE_FEATURE_TEST_MACROS /* for fileno() */
2c6db04a5SJed Brown #include <petscsys.h>                    /*I "petscsys.h" I*/
3f67a399dSBarry Smith #include <petsc/private/petscimpl.h>
4c6db04a5SJed Brown #include <petscconfiginfo.h>
5114011d0SBarry Smith #if defined(PETSC_HAVE_UNISTD_H)
6114011d0SBarry Smith   #include <unistd.h>
7114011d0SBarry Smith #endif
8e5c89e4eSSatish Balay 
9e5c89e4eSSatish Balay /*@C
10aaa8cc7dSPierre Jolivet   PetscIgnoreErrorHandler - Deprecated, use `PetscReturnErrorHandler()`. Ignores the error, allows program to continue as if error did not occur
11e5c89e4eSSatish Balay 
12e5c89e4eSSatish Balay   Not Collective
13e5c89e4eSSatish Balay 
14e5c89e4eSSatish Balay   Input Parameters:
15e32f2f54SBarry Smith + comm - communicator over which error occurred
16e32f2f54SBarry Smith . line - the line number of the error (indicated by __LINE__)
17*10450e9eSJacob Faibussowitsch . fun  - the function name
18e5c89e4eSSatish Balay . file - the file in which the error was detected (indicated by __FILE__)
19e5c89e4eSSatish Balay . mess - an error text string, usually just printed to the screen
20e5c89e4eSSatish Balay . n    - the generic error number
21e5c89e4eSSatish Balay . p    - specific error number
22e5c89e4eSSatish Balay - ctx  - error handler context
23e5c89e4eSSatish Balay 
24e5c89e4eSSatish Balay   Level: developer
25e5c89e4eSSatish Balay 
26811af0c4SBarry Smith   Note:
27811af0c4SBarry Smith   Users do not directly call this routine
28e5c89e4eSSatish Balay 
29db781477SPatrick Sanan .seealso: `PetscReturnErrorHandler()`
30e5c89e4eSSatish Balay  @*/
31d71ae5a4SJacob Faibussowitsch PetscErrorCode PetscIgnoreErrorHandler(MPI_Comm comm, int line, const char *fun, const char *file, PetscErrorCode n, PetscErrorType p, const char *mess, void *ctx)
32d71ae5a4SJacob Faibussowitsch {
33*10450e9eSJacob Faibussowitsch   (void)comm;
34*10450e9eSJacob Faibussowitsch   (void)line;
35*10450e9eSJacob Faibussowitsch   (void)fun;
36*10450e9eSJacob Faibussowitsch   (void)file;
37*10450e9eSJacob Faibussowitsch   (void)p;
38*10450e9eSJacob Faibussowitsch   (void)mess;
39*10450e9eSJacob Faibussowitsch   (void)ctx;
4011cc89d2SBarry Smith   return n;
41e5c89e4eSSatish Balay }
42e5c89e4eSSatish Balay 
43107894f0SSatish Balay /* ---------------------------------------------------------------------------------------*/
44107894f0SSatish Balay 
455abee1b0SJed Brown static char      arch[128], hostname[128], username[128], pname[PETSC_MAX_PATH_LEN], date[128];
46ace3abfcSBarry Smith static PetscBool PetscErrorPrintfInitializeCalled = PETSC_FALSE;
473f6e4ae9SSatish Balay static char      version[256];
48107894f0SSatish Balay 
49107894f0SSatish Balay /*
50107894f0SSatish Balay    Initializes arch, hostname, username, date so that system calls do NOT need
51107894f0SSatish Balay    to be made during the error handler.
52107894f0SSatish Balay */
53d71ae5a4SJacob Faibussowitsch PetscErrorCode PetscErrorPrintfInitialize(void)
54d71ae5a4SJacob Faibussowitsch {
55ace3abfcSBarry Smith   PetscBool use_stdout = PETSC_FALSE, use_none = PETSC_FALSE;
56107894f0SSatish Balay 
57107894f0SSatish Balay   PetscFunctionBegin;
589566063dSJacob Faibussowitsch   PetscCall(PetscGetArchType(arch, sizeof(arch)));
599566063dSJacob Faibussowitsch   PetscCall(PetscGetHostName(hostname, sizeof(hostname)));
609566063dSJacob Faibussowitsch   PetscCall(PetscGetUserName(username, sizeof(username)));
619566063dSJacob Faibussowitsch   PetscCall(PetscGetProgramName(pname, sizeof(pname)));
629566063dSJacob Faibussowitsch   PetscCall(PetscGetDate(date, sizeof(date)));
639566063dSJacob Faibussowitsch   PetscCall(PetscGetVersion(version, sizeof(version)));
64e8fb0fc0SBarry Smith 
659566063dSJacob Faibussowitsch   PetscCall(PetscOptionsGetBool(NULL, NULL, "-error_output_stdout", &use_stdout, NULL));
66a297a907SKarl Rupp   if (use_stdout) PETSC_STDERR = PETSC_STDOUT;
679566063dSJacob Faibussowitsch   PetscCall(PetscOptionsGetBool(NULL, NULL, "-error_output_none", &use_none, NULL));
68a297a907SKarl Rupp   if (use_none) PetscErrorPrintf = PetscErrorPrintfNone;
69107894f0SSatish Balay   PetscErrorPrintfInitializeCalled = PETSC_TRUE;
703ba16761SJacob Faibussowitsch   PetscFunctionReturn(PETSC_SUCCESS);
71107894f0SSatish Balay }
72107894f0SSatish Balay 
73d71ae5a4SJacob Faibussowitsch PetscErrorCode PetscErrorPrintfNone(const char format[], ...)
74d71ae5a4SJacob Faibussowitsch {
753ba16761SJacob Faibussowitsch   return PETSC_SUCCESS;
76e8fb0fc0SBarry Smith }
77e8fb0fc0SBarry Smith 
78d71ae5a4SJacob Faibussowitsch PetscErrorCode PetscErrorPrintfDefault(const char format[], ...)
79d71ae5a4SJacob Faibussowitsch {
80e8fb0fc0SBarry Smith   va_list          Argp;
81ace3abfcSBarry Smith   static PetscBool PetscErrorPrintfCalled = PETSC_FALSE;
823ba16761SJacob Faibussowitsch   PetscErrorCode   ierr;
83e8fb0fc0SBarry Smith 
84e8fb0fc0SBarry Smith   /*
85e8fb0fc0SBarry Smith       This function does not call PetscFunctionBegin and PetscFunctionReturn() because
86e8fb0fc0SBarry Smith     it may be called by PetscStackView().
87e8fb0fc0SBarry Smith 
88e8fb0fc0SBarry Smith       This function does not do error checking because it is called by the error handlers.
89e8fb0fc0SBarry Smith   */
90e8fb0fc0SBarry Smith 
91e8fb0fc0SBarry Smith   if (!PetscErrorPrintfCalled) {
92e8fb0fc0SBarry Smith     PetscErrorPrintfCalled = PETSC_TRUE;
93e8fb0fc0SBarry Smith 
94e8fb0fc0SBarry Smith     /*
95e8fb0fc0SBarry Smith         On the SGI machines and Cray T3E, if errors are generated  "simultaneously" by
96e8fb0fc0SBarry Smith       different processors, the messages are printed all jumbled up; to try to
97e8fb0fc0SBarry Smith       prevent this we have each processor wait based on their rank
98e8fb0fc0SBarry Smith     */
99e8fb0fc0SBarry Smith #if defined(PETSC_CAN_SLEEP_AFTER_ERROR)
100e8fb0fc0SBarry Smith     {
1013ba16761SJacob Faibussowitsch       PetscMPIInt rank = PetscGlobalRank > 8 ? 8 : PetscGlobalRank;
1023ba16761SJacob Faibussowitsch       ierr             = PetscSleep((PetscReal)rank);
1033ba16761SJacob Faibussowitsch       (void)ierr;
104e8fb0fc0SBarry Smith     }
105e8fb0fc0SBarry Smith #endif
106e8fb0fc0SBarry Smith   }
107e8fb0fc0SBarry Smith 
1083ba16761SJacob Faibussowitsch   ierr = PetscFPrintf(PETSC_COMM_SELF, PETSC_STDERR, "[%d]PETSC ERROR: ", PetscGlobalRank);
109e8fb0fc0SBarry Smith   va_start(Argp, format);
1103ba16761SJacob Faibussowitsch   ierr = (*PetscVFPrintf)(PETSC_STDERR, format, Argp);
1113ba16761SJacob Faibussowitsch   (void)ierr;
112e8fb0fc0SBarry Smith   va_end(Argp);
1133ba16761SJacob Faibussowitsch   return PETSC_SUCCESS;
114e8fb0fc0SBarry Smith }
115e8fb0fc0SBarry Smith 
116c2eed0edSBarry Smith /*
117c2eed0edSBarry Smith    On some systems when the stderr is nested through several levels of shell script
118c2eed0edSBarry Smith    before being passed to a file the isatty() falsely returns true resulting in
119c2eed0edSBarry Smith    the screen highlight variables being passed through the test harness. Therefore
120c2eed0edSBarry Smith    simply do not highlight when the PETSC_STDERR is PETSC_STDOUT.
121c2eed0edSBarry Smith */
122d71ae5a4SJacob Faibussowitsch static void PetscErrorPrintfHilight(void)
123d71ae5a4SJacob Faibussowitsch {
12498ed35c3SBarry Smith #if defined(PETSC_HAVE_UNISTD_H) && defined(PETSC_USE_ISATTY)
125c2eed0edSBarry Smith   if (PetscErrorPrintf == PetscErrorPrintfDefault && PETSC_STDERR != PETSC_STDOUT) {
126114011d0SBarry Smith     if (isatty(fileno(PETSC_STDERR))) fprintf(PETSC_STDERR, "\033[1;31m");
127114011d0SBarry Smith   }
128114011d0SBarry Smith #endif
129114011d0SBarry Smith }
130114011d0SBarry Smith 
131d71ae5a4SJacob Faibussowitsch static void PetscErrorPrintfNormal(void)
132d71ae5a4SJacob Faibussowitsch {
13398ed35c3SBarry Smith #if defined(PETSC_HAVE_UNISTD_H) && defined(PETSC_USE_ISATTY)
134c2eed0edSBarry Smith   if (PetscErrorPrintf == PetscErrorPrintfDefault && PETSC_STDERR != PETSC_STDOUT) {
135114011d0SBarry Smith     if (isatty(fileno(PETSC_STDERR))) fprintf(PETSC_STDERR, "\033[0;39m\033[0;49m");
136114011d0SBarry Smith   }
137114011d0SBarry Smith #endif
138114011d0SBarry Smith }
139114011d0SBarry Smith 
14095c0884eSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscOptionsViewError(void);
141114011d0SBarry Smith 
142e5c89e4eSSatish Balay /*@C
143e5c89e4eSSatish Balay 
144e5c89e4eSSatish Balay   PetscTraceBackErrorHandler - Default error handler routine that generates
145e5c89e4eSSatish Balay   a traceback on error detection.
146e5c89e4eSSatish Balay 
147e5c89e4eSSatish Balay   Not Collective
148e5c89e4eSSatish Balay 
149e5c89e4eSSatish Balay   Input Parameters:
150e32f2f54SBarry Smith + comm - communicator over which error occurred
151e32f2f54SBarry Smith . line - the line number of the error (indicated by __LINE__)
152*10450e9eSJacob Faibussowitsch . fun  - the function name
153e5c89e4eSSatish Balay . file - the file in which the error was detected (indicated by __FILE__)
154e5c89e4eSSatish Balay . mess - an error text string, usually just printed to the screen
155e5c89e4eSSatish Balay . n    - the generic error number
156811af0c4SBarry Smith . p    - `PETSC_ERROR_INITIAL` if this is the first call the error handler, otherwise `PETSC_ERROR_REPEAT`
157e5c89e4eSSatish Balay - ctx  - error handler context
158e5c89e4eSSatish Balay 
159811af0c4SBarry Smith   Options Database Keys:
16045b666d6SBarry Smith + -error_output_stdout - output the error messages to stdout instead of the default stderr
16145b666d6SBarry Smith - -error_output_none   - do not output the error messages
162e5c89e4eSSatish Balay 
163e5c89e4eSSatish Balay   Notes:
164811af0c4SBarry Smith   Users do not directly call this routine
165e5c89e4eSSatish Balay 
166811af0c4SBarry Smith   Use `PetscPushErrorHandler()` to set the desired error handler.
167e5c89e4eSSatish Balay 
16845b666d6SBarry Smith   Level: developer
169e5c89e4eSSatish Balay 
170db781477SPatrick Sanan .seealso: `PetscError()`, `PetscPushErrorHandler()`, `PetscPopErrorHandler()`, `PetscAttachDebuggerErrorHandler()`,
171db781477SPatrick Sanan           `PetscAbortErrorHandler()`, `PetscMPIAbortErrorHandler()`, `PetscReturnErrorHandler()`, `PetscEmacsClientErrorHandler()`
172e5c89e4eSSatish Balay  @*/
173d71ae5a4SJacob Faibussowitsch PetscErrorCode PetscTraceBackErrorHandler(MPI_Comm comm, int line, const char *fun, const char *file, PetscErrorCode n, PetscErrorType p, const char *mess, void *ctx)
174d71ae5a4SJacob Faibussowitsch {
1753ba16761SJacob Faibussowitsch   PetscErrorCode ierr;
176e5c89e4eSSatish Balay   PetscLogDouble mem, rss;
177574034a9SJed Brown   PetscBool      flg1 = PETSC_FALSE, flg2 = PETSC_FALSE, flg3 = PETSC_FALSE;
178997adca8SBarry Smith   PetscMPIInt    rank = 0;
179e5c89e4eSSatish Balay 
180*10450e9eSJacob Faibussowitsch   (void)ctx;
181a297a907SKarl Rupp   if (comm != PETSC_COMM_SELF) MPI_Comm_rank(comm, &rank);
182a297a907SKarl Rupp 
183660278c0SBarry Smith   if (rank == 0 && (!PetscCIEnabledPortableErrorOutput || PetscGlobalRank == 0)) {
184114011d0SBarry Smith     static int cnt         = 1;
18515858b40SBarry Smith     static int repeaterror = 0;
186114011d0SBarry Smith 
18715858b40SBarry Smith     if (p == PETSC_ERROR_INITIAL && cnt > 1) {
18815858b40SBarry Smith       if (repeaterror < 1) {
18915858b40SBarry Smith         ierr = (*PetscErrorPrintf)("\n");
19015858b40SBarry Smith         ierr = (*PetscErrorPrintf)("  It appears a new error in the code was triggered after a previous error was not properly handled\n");
19115858b40SBarry Smith         ierr = (*PetscErrorPrintf)("  via (for example) the use of PetscCall(TheFunctionThatErrors());\n");
19215858b40SBarry Smith         ierr = (*PetscErrorPrintf)("  Above is the traceback for the previous unhandled error, below the traceback for the next error\n");
19315858b40SBarry Smith         ierr = (*PetscErrorPrintf)("  ALL ERRORS in the PETSc libraries are fatal, you should add the appropriate error checking to the code\n");
19415858b40SBarry Smith         ierr = (*PetscErrorPrintf)("\n");
19515858b40SBarry Smith         cnt  = 1;
19615858b40SBarry Smith         repeaterror++;
19715858b40SBarry Smith       } else {
19815858b40SBarry Smith         ierr = (*PetscErrorPrintf)("\n");
19915858b40SBarry Smith         ierr = (*PetscErrorPrintf)("  There are multiple unhandled errors in the code; aborting program\n");
20015858b40SBarry Smith         ierr = (*PetscErrorPrintf)("\n");
20115858b40SBarry Smith         ierr = PetscOptionsViewError();
20215858b40SBarry Smith         PetscErrorPrintfHilight();
20315858b40SBarry Smith         ierr = (*PetscErrorPrintf)("----------------End of Error Message -------send entire error message to petsc-maint@mcs.anl.gov----------\n");
20415858b40SBarry Smith         PetscErrorPrintfNormal();
20515858b40SBarry Smith         PETSCABORT(comm, ierr);
20615858b40SBarry Smith       }
20715858b40SBarry Smith     }
20857a177adSBarry Smith     if (cnt == 1) {
209114011d0SBarry Smith       PetscErrorPrintfHilight();
2103ba16761SJacob Faibussowitsch       ierr = (*PetscErrorPrintf)("--------------------- Error Message --------------------------------------------------------------\n");
2115f3a2c8dSBarry Smith       PetscErrorPrintfNormal();
212e5c89e4eSSatish Balay       if (n == PETSC_ERR_MEM) {
2133ba16761SJacob Faibussowitsch         ierr = (*PetscErrorPrintf)("Out of memory. This could be due to allocating\n");
2143ba16761SJacob Faibussowitsch         ierr = (*PetscErrorPrintf)("too large an object or bleeding by not properly\n");
2153ba16761SJacob Faibussowitsch         ierr = (*PetscErrorPrintf)("destroying unneeded objects.\n");
2163ba16761SJacob Faibussowitsch         ierr = PetscMallocGetCurrentUsage(&mem);
2173ba16761SJacob Faibussowitsch         ierr = PetscMemoryGetCurrentUsage(&rss);
2183ba16761SJacob Faibussowitsch         ierr = PetscOptionsGetBool(NULL, NULL, "-malloc_dump", &flg1, NULL);
2193ba16761SJacob Faibussowitsch         ierr = PetscOptionsGetBool(NULL, NULL, "-malloc_view", &flg2, NULL);
2203ba16761SJacob Faibussowitsch         ierr = PetscOptionsHasName(NULL, NULL, "-malloc_view_threshold", &flg3);
2213ba16761SJacob Faibussowitsch         if (flg2 || flg3) ierr = PetscMallocView(stdout);
222a297a907SKarl Rupp         else {
2233ba16761SJacob Faibussowitsch           ierr = (*PetscErrorPrintf)("Memory allocated %.0f Memory used by process %.0f\n", mem, rss);
2243ba16761SJacob Faibussowitsch           if (flg1) ierr = PetscMallocDump(stdout);
2253ba16761SJacob Faibussowitsch           else ierr = (*PetscErrorPrintf)("Try running with -malloc_dump or -malloc_view for info.\n");
226e5c89e4eSSatish Balay         }
227e5c89e4eSSatish Balay       } else {
228e5c89e4eSSatish Balay         const char *text;
2293ba16761SJacob Faibussowitsch         ierr = PetscErrorMessage(n, &text, NULL);
2303ba16761SJacob Faibussowitsch         if (text) ierr = (*PetscErrorPrintf)("%s\n", text);
231e5c89e4eSSatish Balay       }
2323ba16761SJacob Faibussowitsch       if (mess) ierr = (*PetscErrorPrintf)("%s\n", mess);
2333ba16761SJacob Faibussowitsch       ierr = PetscOptionsLeftError();
2343ba16761SJacob Faibussowitsch       ierr = (*PetscErrorPrintf)("See https://petsc.org/release/faq/ for trouble shooting.\n");
235660278c0SBarry Smith       if (!PetscCIEnabledPortableErrorOutput) {
2363ba16761SJacob Faibussowitsch         ierr = (*PetscErrorPrintf)("%s\n", version);
2373ba16761SJacob Faibussowitsch         if (PetscErrorPrintfInitializeCalled) ierr = (*PetscErrorPrintf)("%s on a %s named %s by %s %s\n", pname, arch, hostname, username, date);
2383ba16761SJacob Faibussowitsch         ierr = (*PetscErrorPrintf)("Configure options %s\n", petscconfigureoptions);
239107894f0SSatish Balay       }
240660278c0SBarry Smith     }
241997adca8SBarry Smith     /* print line of stack trace */
2423ba16761SJacob Faibussowitsch     if (fun) ierr = (*PetscErrorPrintf)("#%d %s() at %s:%d\n", cnt++, fun, PetscCIFilename(file), PetscCILinenumber(line));
2433ba16761SJacob Faibussowitsch     else if (file) ierr = (*PetscErrorPrintf)("#%d %s:%d\n", cnt++, PetscCIFilename(file), PetscCILinenumber(line));
24449c86fc7SBarry Smith     if (fun) {
245bbcf679cSJacob Faibussowitsch       PetscBool ismain = PETSC_FALSE;
246bbcf679cSJacob Faibussowitsch 
2473ba16761SJacob Faibussowitsch       ierr = PetscStrncmp(fun, "main", 4, &ismain);
248fbfcfee5SBarry Smith       if (ismain) {
2493ba16761SJacob Faibussowitsch         if ((n <= PETSC_ERR_MIN_VALUE) || (n >= PETSC_ERR_MAX_VALUE)) ierr = (*PetscErrorPrintf)("Reached the main program with an out-of-range error code %d. This should never happen\n", n);
2503ba16761SJacob Faibussowitsch         ierr = PetscOptionsViewError();
251114011d0SBarry Smith         PetscErrorPrintfHilight();
2523ba16761SJacob Faibussowitsch         ierr = (*PetscErrorPrintf)("----------------End of Error Message -------send entire error message to petsc-maint@mcs.anl.gov----------\n");
253114011d0SBarry Smith         PetscErrorPrintfNormal();
254114011d0SBarry Smith       }
25549c86fc7SBarry Smith     }
256997adca8SBarry Smith   } else {
257a5b23f4aSJose E. Roman     /* do not print error messages since process 0 will print them, sleep before aborting so will not accidentally kill process 0*/
2583ba16761SJacob Faibussowitsch     ierr = PetscSleep(10.0);
259c16385d5SBarry Smith     exit(0);
260997adca8SBarry Smith   }
2613ba16761SJacob Faibussowitsch   (void)ierr;
262362febeeSStefano Zampini   return n;
263e5c89e4eSSatish Balay }
264