I've been using libzip to handle zip files, and have based my code on the example found in rodrigo's answer to this question. Here's his code, for quick reference:
#include <zip.h>
int main()
{
//Open the ZIP archive
int err = 0;
zip *z = zip_open("foo.zip", 0, &err);
//Search for the file of given name
const char *name = "file.txt";
struct zip_stat st;
zip_stat_init(&st);
zip_stat(z, name, 0, &st);
//Alloc memory for its uncompressed contents
char *contents = new char[st.size];
//Read the compressed file
zip_file *f = zip_fopen(z, "file.txt", 0);
zip_fread(f, contents, st.size);
zip_fclose(f);
//And close the archive
zip_close(z);
}
I traced the errors I subsequently got from Valgrind back to this code- it complains of an uninitialised value when opening the zipped "file.txt" using 'zip_fopen()`.
==29256== Conditional jump or move depends on uninitialised value(s)
==29256== at 0x5B4B290: inflateReset2 (in /usr/lib/libz.so.1.2.3.4)
==29256== by 0x5B4B37F: inflateInit2_ (in /usr/lib/libz.so.1.2.3.4)
==29256== by 0x4E2EB8C: zip_fopen_index (in /usr/lib/libzip.so.1.0.0)
==29256== by 0x400C32: main (main.cpp:24)
==29256== Uninitialised value was created by a heap allocation
==29256== at 0x4C244E8: malloc (vg_replace_malloc.c:236)
==29256== by 0x5B4B35B: inflateInit2_ (in /usr/lib/libz.so.1.2.3.4)
==29256== by 0x4E2EB8C: zip_fopen_index (in /usr/lib/libzip.so.1.0.0)
==29256== by 0x400C32: main (main.cpp:24)
==29256==
==29256==
==29256== HEAP SUMMARY:
==29256== in use at exit: 71 bytes in 1 blocks
==29256== total heap usage: 26 allocs, 25 frees, 85,851 bytes allocated
==29256==
==29256== 71 bytes in 1 blocks are definitely lost in loss record 1 of 1
==29256== at 0x4C24A72: operator new[](unsigned long) (vg_replace_malloc.c:305)
==29256== by 0x400BEE: main (main.cpp:19)
I can't see where the uninitialised value is coming from in this code. Can anyone trace this, or does the fault lie with libzip itself? Should I switch to another zip library- for example, Minizip?
EDIT: The 71 bytes are the contents of file.txt which was read in- delete[] contents;
tagged on the end will eliminate that.
(I'd have left a comment on the original answer to draw attention to the issue, but I do not have the requisite rep.)