cd regress/build
make
cd patches/testing
- g++ -g -Wall -I../../src -I../../src/lib -L../../src/lib justdisk.c -lbac -lpthread -lssl -D_TEST_BUF
+ g++ -g -Wall -I../../src -I../../src/lib -L../../src/lib justdisk.c -lbac -lpthread -lssl -D_TEST_TREE
+ find / > lst
./a.out
*/
#include <string.h>
#include <fcntl.h>
#include "bacula.h"
-
+#include "dird/dird.h"
typedef struct {
- int32_t path;
- int32_t filename;
+ char *path;
+ char *fname;
int32_t seen;
int32_t pad;
int64_t mtime;
int64_t ctime;
} AccurateElt;
-#define NB_ELT 50000000
+#define NB_ELT 500
typedef struct {
char *buf;
- rblink *link;
+ rblink link;
} MY;
+typedef struct {
+ char *path;
+ char *fname;
+ int64_t index;
+ rblink link;
+} MY_INDEX;
+
+static int my_index_cmp(void *item1, void *item2)
+{
+ MY_INDEX *elt1, *elt2;
+ elt1 = (MY_INDEX *) item1;
+ elt2 = (MY_INDEX *) item2;
+ int r = strcmp(elt1->fname, elt2->fname);
+ if (r) {
+ return r;
+ }
+ r = strcmp(elt1->path, elt2->path);
+ return r;
+}
+
static int my_cmp(void *item1, void *item2)
{
MY *elt1, *elt2;
elt1 = (MY *) item1;
elt2 = (MY *) item2;
- printf("cmp(%s, %s)\n", elt1->buf, elt2->buf);
return strcmp(elt1->buf, elt2->buf);
}
char buffer[1024];
MY *res;
rblist *lst;
- lst = New(rblist(res, res->link));
+ lst = New(rblist(res, &res->link));
fp = fopen(file, "r");
if (!fp) {
_dirty = 1;
}
+typedef struct B_DBx
+{
+ POOLMEM *fname;
+ int fnl;
+ POOLMEM *path;
+ int pnl;
+} B_DBx;
+
+/*
+ * Given a full filename, split it into its path
+ * and filename parts. They are returned in pool memory
+ * in the mdb structure.
+ */
+void split_path_and_file(B_DBx *mdb, const char *fname)
+{
+ const char *p, *f;
+
+ /* Find path without the filename.
+ * I.e. everything after the last / is a "filename".
+ * OK, maybe it is a directory name, but we treat it like
+ * a filename. If we don't find a / then the whole name
+ * must be a path name (e.g. c:).
+ */
+ for (p=f=fname; *p; p++) {
+ if (IsPathSeparator(*p)) {
+ f = p; /* set pos of last slash */
+ }
+ }
+ if (IsPathSeparator(*f)) { /* did we find a slash? */
+ f++; /* yes, point to filename */
+ } else { /* no, whole thing must be path name */
+ f = p;
+ }
+
+ /* If filename doesn't exist (i.e. root directory), we
+ * simply create a blank name consisting of a single
+ * space. This makes handling zero length filenames
+ * easier.
+ */
+ mdb->fnl = p - f;
+ if (mdb->fnl > 0) {
+ mdb->fname = check_pool_memory_size(mdb->fname, mdb->fnl+1);
+ memcpy(mdb->fname, f, mdb->fnl); /* copy filename */
+ mdb->fname[mdb->fnl] = 0;
+ } else {
+ mdb->fname[0] = 0;
+ mdb->fnl = 0;
+ }
+
+ mdb->pnl = f - fname;
+ if (mdb->pnl > 0) {
+ mdb->path = check_pool_memory_size(mdb->path, mdb->pnl+1);
+ memcpy(mdb->path, fname, mdb->pnl);
+ mdb->path[mdb->pnl] = 0;
+ } else {
+ mdb->path[0] = 0;
+ mdb->pnl = 0;
+ }
+
+ Dmsg2(500, "split path=%s file=%s\n", mdb->path, mdb->fname);
+}
+
+#ifdef _TEST_TREE
+int main()
+{
+ AccurateElt *elt;
+ char *start_heap = (char *)sbrk(0);
+ MY *myp, *res;
+ MY_INDEX *idx;
+
+ FILE *fp;
+ fp = fopen("lst", "r");
+ if (!fp) {
+ exit (5);
+ }
+ rblist *index, *rb_file, *rb_path;
+ index = New(rblist(idx, &idx->link)); /* (pathid, filenameid) => index */
+ rb_path = New(rblist(myp, &myp->link)); /* pathid => path */
+ rb_file = New(rblist(myp, &myp->link)); /* filenameid => filename */
+
+ B_DBx mdb;
+ mdb.fname = get_pool_memory(PM_FNAME);
+ mdb.path = get_pool_memory(PM_FNAME);
+
+ AccurateBuffer *accbuf = new AccurateBuffer;
+ char buf[4096];
+ int64_t count=0;
+ printf("loading...\n");
+ while (fgets(buf, sizeof(buf), fp)) {
+ int len = strlen(buf);
+ if (len > 0) {
+ buf[len-1]=0; /* zap \n */
+ }
+ split_path_and_file(&mdb, buf);
+
+ MY_INDEX *idx = (MY_INDEX *)malloc(sizeof(MY_INDEX));
+ memset(idx, 0, sizeof(MY_INDEX));
+
+ myp = (MY *)malloc(sizeof(MY));
+ memset(myp, 0, sizeof(MY));
+ myp->buf = bstrdup(mdb.path);
+ res = (MY *)rb_path->insert(myp, my_cmp);
+ if (res != myp) {
+ free(myp->buf);
+ free(myp);
+ }
+
+ idx->path = res->buf;
+
+ myp = (MY *)malloc(sizeof(MY));
+ memset(myp, 0, sizeof(MY));
+ myp->buf = bstrdup(mdb.fname);
+ res = (MY *)rb_file->insert(myp, my_cmp);
+ if (res != myp) {
+ free(myp->buf);
+ free(myp);
+ }
+
+ idx->fname = res->buf;
+
+ idx->index = count;
+ index->insert(idx, my_index_cmp);
+
+ elt = (AccurateElt *)accbuf->get_elt(count);
+ elt->mtime = count;
+ elt->path = idx->path;
+ elt->fname = idx->fname;
+ accbuf->update_elt(count);
+
+ count++;
+ }
+
+ printf("fetch and update...\n");
+ fprintf(stderr, "count;%lli\n", count);
+ fprintf(stderr, "heap;%lld\n", (long long)((char *)sbrk(0) - start_heap));
+
+ fseek(fp, 0, SEEK_SET);
+ int toggle=0;
+ while (fgets(buf, sizeof(buf), fp)) {
+ int len = strlen(buf);
+ if (len > 0) {
+ buf[len-1]=0; /* zap \n */
+ }
+ split_path_and_file(&mdb, buf);
+
+ MY search;
+ search.buf = mdb.fname;
+ MY *f = (MY *)rb_file->search(&search, my_cmp);
+
+ if (!f) {
+ continue; /* not found, skip */
+ }
+ search.buf = mdb.path;
+ MY *p = (MY *)rb_path->search(&search, my_cmp);
+ if (!p) {
+ continue; /* not found, skip */
+ }
+
+ MY_INDEX idx;
+ idx.path = p->buf;
+ idx.fname = f->buf;
+ MY_INDEX *res = (MY_INDEX *)index->search(&idx, my_index_cmp);
+
+ if (!res) {
+ continue; /* not found skip */
+ }
+
+ AccurateElt *elt = (AccurateElt *)accbuf->get_elt(res->index);
+ elt->seen=toggle;
+ accbuf->update_elt(res->index);
+ toggle = (toggle+1)%100000;
+ }
+
+ for(int j=0;j<=count; j++) {
+ AccurateElt *elt = (AccurateElt *)accbuf->get_elt(j);
+ if (elt->path && elt->fname) {
+ if (!elt->seen) {
+ printf("%s%s\n", elt->path, elt->fname);
+ }
+ }
+ }
+}
+#else
#ifdef _TEST_OPEN
int main()
{
AccurateElt elt;
char *start_heap = (char *)sbrk(0);
- rblist *rb_file = load_rb("files");
- rblist *rb_path = load_rb("path");
-
- char *etc = (char *)rb_path->search((void *)"/etc/", my_cmp);
-
- printf("%p\n", etc);
-
fd = open("testfile", O_CREAT | O_RDWR, 0600);
if (fd<0) {
perror("E: Can't open testfile ");
}
#endif /* _TEST_BUF */
#endif /* _TEST_OPEN */
+#endif /* _TEST_TREE */
projects / bacula / bacula / blobdiff