fclose() causing segmentation fault
-
22-07-2019 - |
Question
I have a tab-delimited text file that I am parsing. Its first column contains strings of the format chrX
, where X
denotes a set of strings, e.g., "1", "2", ..., "X", "Y".
These are each stored in a char*
called chromosome
, as the file is parsed.
The text file is sorted on the first column lexicographically, i.e., I will have a number of rows starting with "chr1", and then "chr2", etc.
At each "chrX" entry, I need to open another file that is associated with this entry:
FILE *merbaseIn;
// loop through rows...
if (chromosome == NULL)
openSourceFile(&chromosome, fieldArray[i], &merbaseIn, GENPATHIN);
else {
if (strcmp(chromosome, fieldArray[i]) != 0) { // new chromosome
fclose(merbaseIn); // close old chromosome FILE ptr
free(chromosome); // free old chromosome ptr
openSourceFile(&chromosome, fieldArray[i], &merbaseIn, GENPATHIN); // set up new chromosome FILE ptr
}
}
// parse row
I have the function openSourceFile
that is defined as follows:
void openSourceFile (char** chrome, const char* field, FILE** filePtr, const char *path) {
char filename[100];
*chrome = (char *) malloc ((size_t) strlen(field));
if (*chrome == NULL) {
fprintf(stderr, "ERROR: Cannot allocate memory for chromosome name!");
exit(EXIT_FAILURE);
}
strcpy(*chrome, field);
sprintf(filename,"%s%s.fa", path, field);
*filePtr = fopen(filename, "r");
if (*filePtr == NULL) {
fprintf(stderr, "ERROR: Could not open fasta source file %s\n", filename);
exit(EXIT_FAILURE);
}
}
The problem is that my application quits with a Segmentation Fault going from the first chromosome to the second (from chr1
to chr2
) at the following line, where I close the first chromosome file that I opened:
fclose(merbaseIn);
I know I'm not passing fclose
a NULL pointer, because up until the Segmentation Fault, I am reading data from this file. I can even wrap this in a conditional and I still get the Fault:
if (merbaseIn != NULL) {
fclose(merbaseIn);
}
Further, I know openSourceFile
works (at least for chr1
, when setting up the first file handle of FILE*
) because my application parses chr1
rows and reads data from the FILE*
source file correctly.
What is it about this fclose
call that is causing a Segmentation Fault to occur?
Solution
valgrind --db-attach=yes --leak-check=yes --tool=memcheck --num-callers=16 --leak-resolution=high ./yourprogram args
It's very likely the segfault is caused by memory corruption on the heap, not anything that's affecting locals. Valgrind will immediately show you the first wrong access you make.
Edit: The --db-attach
option to valgrind
has been deprecated since release 3.10.0 in 2014. The release notes state:
The built-in GDB server capabilities are superior and should be used
instead. Learn more here:
http://valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.gdbserver
OTHER TIPS
one error i notice is this line:
*chrome = (char *) malloc ((size_t) strlen(field));
which should be:
*chrome = (char *) malloc ((size_t) strlen(field)+1);
This is because a string has a closing 0 at the end which you also have to make room for
Best guess is that some other part of your code is corrupting memory through a buffer overrun or similar bug.
Although unlikely to be the cause, you have a potential overrun condition in your filename array when the full filename exceeds 100 chars.
I would suggest using a debugger to watch for changes in the memory location used by the merbaseIn variable.
Generic Pointer Problem
C is a great language but it does require that you not clobber your own memory. Besides the problem previously noted where the malloc is 1 byte short, you may have other pointer problems.
I would suggest using a memory debugger. In the past, Electric Fence was rather popular but these days I hear more about valgrind. There are lots of other choices.
In addition to error found by reinier, I suspect that:
free(chromosome);
should be followed by:
chromosome = NULL;
in order to prevent potential usage of a no longer valid value.
why this FILE** filePtr if only this FILE* filePtr would be enough ? just a thought...
valgrind
with memcheck
is definitely the right tool to discover the cause of the segmentation fault. To use a debugger with valgrind
, note that the --db-attach
option to valgrind
has been deprecated since Valgrind 3.10.0 was released in 2014. The release notes state:
The built-in GDB server capabilities are superior and should be used
instead. Learn more here:
http://valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.gdbserver
Review every place you used "malloc" and see if you made a mistake.
For example, I was putting lines read from a file into a char**, but I incorrectly malloced it as:
my_list = malloc(sizeof(char) * num_lines_found_in_file);
When it should have been:
my_list = malloc(sizeof(char*)* num_lines_found_in_file);