diff --git a/src/gnudiff_analyze.cpp b/src/gnudiff_analyze.cpp index 469a80c..b6616b6 100644 --- a/src/gnudiff_analyze.cpp +++ b/src/gnudiff_analyze.cpp @@ -1,847 +1,847 @@ /* Analyze file differences for GNU DIFF. Modified for KDiff3 by Joachim Eibl 2003. The original file was part of GNU DIFF. Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1998, 2001, 2002 Free Software Foundation, Inc. GNU DIFF is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version. GNU DIFF is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; see the file COPYING. If not, write to the Free Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ /* The basic algorithm is described in: "An O(ND) Difference Algorithm and its Variations", Eugene Myers, Algorithmica Vol. 1 No. 2, 1986, pp. 251-266; see especially section 4.2, which describes the variation used below. Unless the --minimal option is specified, this code uses the TOO_EXPENSIVE heuristic, by Paul Eggert, to limit the cost to O(N**1.5 log N) at the price of producing suboptimal output for large inputs with many differences. The basic algorithm was independently discovered as described in: "Algorithms for Approximate String Matching", E. Ukkonen, Information and Control Vol. 64, 1985, pp. 100-118. */ #define GDIFF_MAIN #include "gnudiff_diff.h" #include // for max, min #include static GNULineRef *xvec, *yvec; /* Vectors being compared. */ static GNULineRef *fdiag; /* Vector, indexed by diagonal, containing 1 + the X coordinate of the point furthest along the given diagonal in the forward search of the edit matrix. */ static GNULineRef *bdiag; /* Vector, indexed by diagonal, containing the X coordinate of the point furthest along the given diagonal in the backward search of the edit matrix. */ static GNULineRef too_expensive; /* Edit scripts longer than this are too expensive to compute. */ #define SNAKE_LIMIT 20 /* Snakes bigger than this are considered `big'. */ struct partition { GNULineRef xmid, ymid; /* Midpoints of this partition. */ bool lo_minimal; /* Nonzero if low half will be analyzed minimally. */ bool hi_minimal; /* Likewise for high half. */ }; /* Find the midpoint of the shortest edit script for a specified portion of the two files. Scan from the beginnings of the files, and simultaneously from the ends, doing a breadth-first search through the space of edit-sequence. When the two searches meet, we have found the midpoint of the shortest edit sequence. If FIND_MINIMAL is nonzero, find the minimal edit script regardless of expense. Otherwise, if the search is too expensive, use heuristics to stop the search and report a suboptimal answer. Set PART->(xmid,ymid) to the midpoint (XMID,YMID). The diagonal number XMID - YMID equals the number of inserted lines minus the number of deleted lines (counting only lines before the midpoint). Return the approximate edit cost; this is the total number of lines inserted or deleted (counting only lines before the midpoint), unless a heuristic is used to terminate the search prematurely. Set PART->lo_minimal to true iff the minimal edit script for the left half of the partition is known; similarly for PART->hi_minimal. This function assumes that the first lines of the specified portions of the two files do not match, and likewise that the last lines do not match. The caller must trim matching lines from the beginning and end of the portions it is going to specify. If we return the "wrong" partitions, the worst this can do is cause suboptimal diff output. It cannot cause incorrect diff output. */ GNULineRef GnuDiff::diag(GNULineRef xoff, GNULineRef xlim, GNULineRef yoff, GNULineRef ylim, bool find_minimal, partition *part) { GNULineRef *const fd = fdiag; /* Give the compiler a chance. */ GNULineRef *const bd = bdiag; /* Additional help for the compiler. */ GNULineRef const *const xv = xvec; /* Still more help for the compiler. */ GNULineRef const *const yv = yvec; /* And more and more . . . */ GNULineRef const dmin = xoff - ylim; /* Minimum valid diagonal. */ GNULineRef const dmax = xlim - yoff; /* Maximum valid diagonal. */ GNULineRef const fmid = xoff - yoff; /* Center diagonal of top-down search. */ GNULineRef const bmid = xlim - ylim; /* Center diagonal of bottom-up search. */ GNULineRef fmin = fmid, fmax = fmid; /* Limits of top-down search. */ GNULineRef bmin = bmid, bmax = bmid; /* Limits of bottom-up search. */ GNULineRef c; /* Cost. */ bool odd = (fmid - bmid) & 1; /* True if southeast corner is on an odd diagonal with respect to the northwest. */ fd[fmid] = xoff; bd[bmid] = xlim; for(c = 1;; ++c) { GNULineRef d; /* Active diagonal. */ bool big_snake = false; /* Extend the top-down search by an edit step in each diagonal. */ fmin > dmin ? fd[--fmin - 1] = -1 : ++fmin; fmax < dmax ? fd[++fmax + 1] = -1 : --fmax; for(d = fmax; d >= fmin; d -= 2) { GNULineRef x, y, oldx, tlo = fd[d - 1], thi = fd[d + 1]; if(tlo >= thi) x = tlo + 1; else x = thi; oldx = x; y = x - d; while(x < xlim && y < ylim && xv[x] == yv[y]) ++x, ++y; if(x - oldx > SNAKE_LIMIT) big_snake = true; fd[d] = x; if(odd && bmin <= d && d <= bmax && bd[d] <= x) { part->xmid = x; part->ymid = y; part->lo_minimal = part->hi_minimal = true; return 2 * c - 1; } } /* Similarly extend the bottom-up search. */ bmin > dmin ? bd[--bmin - 1] = GNULINEREF_MAX : ++bmin; bmax < dmax ? bd[++bmax + 1] = GNULINEREF_MAX : --bmax; for(d = bmax; d >= bmin; d -= 2) { GNULineRef x, y, oldx, tlo = bd[d - 1], thi = bd[d + 1]; if(tlo < thi) x = tlo; else x = thi - 1; oldx = x; y = x - d; while(x > xoff && y > yoff && xv[x - 1] == yv[y - 1]) --x, --y; if(oldx - x > SNAKE_LIMIT) big_snake = true; bd[d] = x; if(!odd && fmin <= d && d <= fmax && x <= fd[d]) { part->xmid = x; part->ymid = y; part->lo_minimal = part->hi_minimal = true; return 2 * c; } } if(find_minimal) continue; /* Heuristic: check occasionally for a diagonal that has made lots of progress compared with the edit distance. If we have any such, find the one that has made the most progress and return it as if it had succeeded. With this heuristic, for files with a constant small density of changes, the algorithm is linear in the file size. */ if(200 < c && big_snake && speed_large_files) { GNULineRef best; best = 0; for(d = fmax; d >= fmin; d -= 2) { GNULineRef dd = d - fmid; GNULineRef x = fd[d]; GNULineRef y = x - d; GNULineRef v = (x - xoff) * 2 - dd; if(v > 12 * (c + (dd < 0 ? -dd : dd))) { if(v > best && xoff + SNAKE_LIMIT <= x && x < xlim && yoff + SNAKE_LIMIT <= y && y < ylim) { /* We have a good enough best diagonal; now insist that it end with a significant snake. */ int k; for(k = 1; xv[x - k] == yv[y - k]; k++) if(k == SNAKE_LIMIT) { best = v; part->xmid = x; part->ymid = y; break; } } } } if(best > 0) { part->lo_minimal = true; part->hi_minimal = false; return 2 * c - 1; } best = 0; for(d = bmax; d >= bmin; d -= 2) { GNULineRef dd = d - bmid; GNULineRef x = bd[d]; GNULineRef y = x - d; GNULineRef v = (xlim - x) * 2 + dd; if(v > 12 * (c + (dd < 0 ? -dd : dd))) { if(v > best && xoff < x && x <= xlim - SNAKE_LIMIT && yoff < y && y <= ylim - SNAKE_LIMIT) { /* We have a good enough best diagonal; now insist that it end with a significant snake. */ int k; for(k = 0; xv[x + k] == yv[y + k]; k++) if(k == SNAKE_LIMIT - 1) { best = v; part->xmid = x; part->ymid = y; break; } } } } if(best > 0) { part->lo_minimal = false; part->hi_minimal = true; return 2 * c - 1; } } /* Heuristic: if we've gone well beyond the call of duty, give up and report halfway between our best results so far. */ if(c >= too_expensive) { GNULineRef fxybest, fxbest; GNULineRef bxybest, bxbest; fxbest = bxbest = 0; /* Pacify `gcc -Wall'. */ /* Find forward diagonal that maximizes X + Y. */ fxybest = -1; for(d = fmax; d >= fmin; d -= 2) { GNULineRef x = std::min(fd[d], xlim); GNULineRef y = x - d; if(ylim < y) x = ylim + d, y = ylim; if(fxybest < x + y) { fxybest = x + y; fxbest = x; } } /* Find backward diagonal that minimizes X + Y. */ bxybest = GNULINEREF_MAX; for(d = bmax; d >= bmin; d -= 2) { GNULineRef x = std::max(xoff, bd[d]); GNULineRef y = x - d; if(y < yoff) x = yoff + d, y = yoff; if(x + y < bxybest) { bxybest = x + y; bxbest = x; } } /* Use the better of the two diagonals. */ if((xlim + ylim) - bxybest < fxybest - (xoff + yoff)) { part->xmid = fxbest; part->ymid = fxybest - fxbest; part->lo_minimal = true; part->hi_minimal = false; } else { part->xmid = bxbest; part->ymid = bxybest - bxbest; part->lo_minimal = false; part->hi_minimal = true; } return 2 * c - 1; } } } /* Compare in detail contiguous subsequences of the two files which are known, as a whole, to match each other. The results are recorded in the vectors files[N].changed, by storing 1 in the element for each line that is an insertion or deletion. The subsequence of file 0 is [XOFF, XLIM) and likewise for file 1. Note that XLIM, YLIM are exclusive bounds. All line numbers are origin-0 and discarded lines are not counted. If FIND_MINIMAL, find a minimal difference no matter how expensive it is. */ void GnuDiff::compareseq(GNULineRef xoff, GNULineRef xlim, GNULineRef yoff, GNULineRef ylim, bool find_minimal) { GNULineRef *const xv = xvec; /* Help the compiler. */ GNULineRef *const yv = yvec; /* Slide down the bottom initial diagonal. */ while(xoff < xlim && yoff < ylim && xv[xoff] == yv[yoff]) ++xoff, ++yoff; /* Slide up the top initial diagonal. */ while(xlim > xoff && ylim > yoff && xv[xlim - 1] == yv[ylim - 1]) --xlim, --ylim; /* Handle simple cases. */ if(xoff == xlim) while(yoff < ylim) files[1].changed[files[1].realindexes[yoff++]] = true; else if(yoff == ylim) while(xoff < xlim) files[0].changed[files[0].realindexes[xoff++]] = true; else { GNULineRef c; partition part; /* Find a point of correspondence in the middle of the files. */ c = diag(xoff, xlim, yoff, ylim, find_minimal, &part); /* This should be impossible, because it implies that one of the two subsequences is empty, and that case was handled above without calling `diag'. */ Q_ASSERT(c != 1); /* Use the partitions to split this problem into subproblems. */ compareseq(xoff, part.xmid, yoff, part.ymid, part.lo_minimal); compareseq(part.xmid, xlim, part.ymid, ylim, part.hi_minimal); } } /* Discard lines from one file that have no matches in the other file. A line which is discarded will not be considered by the actual comparison algorithm; it will be as if that line were not in the file. The file's `realindexes' table maps virtual line numbers (which don't count the discarded lines) into real line numbers; this is how the actual comparison algorithm produces results that are comprehensible when the discarded lines are counted. When we discard a line, we also mark it as a deletion or insertion so that it will be printed in the output. */ void GnuDiff::discard_confusing_lines(file_data filevec[]) { int f; GNULineRef i; char *discarded[2]; GNULineRef *equiv_count[2]; GNULineRef *p; /* Allocate our results. */ - p = (GNULineRef *)xmalloc((filevec[0].buffered_lines + filevec[1].buffered_lines) * (2 * sizeof *p)); + p = (GNULineRef *)xmalloc((filevec[0].buffered_lines + filevec[1].buffered_lines) * (2 * sizeof(*p))); for(f = 0; f < 2; ++f) { filevec[f].undiscarded = p; p += filevec[f].buffered_lines; filevec[f].realindexes = p; p += filevec[f].buffered_lines; } /* Set up equiv_count[F][I] as the number of lines in file F that fall in equivalence class I. */ - p = (GNULineRef *)zalloc(filevec[0].equiv_max * (2 * sizeof *p)); + p = (GNULineRef *)zalloc(filevec[0].equiv_max * (2 * sizeof(*p))); equiv_count[0] = p; equiv_count[1] = p + filevec[0].equiv_max; for(i = 0; i < filevec[0].buffered_lines; ++i) ++equiv_count[0][filevec[0].equivs[i]]; for(i = 0; i < filevec[1].buffered_lines; ++i) ++equiv_count[1][filevec[1].equivs[i]]; /* Set up tables of which lines are going to be discarded. */ discarded[0] = (char *)zalloc(filevec[0].buffered_lines + filevec[1].buffered_lines); discarded[1] = discarded[0] + filevec[0].buffered_lines; /* Mark to be discarded each line that matches no line of the other file. If a line matches many lines, mark it as provisionally discardable. */ for(f = 0; f < 2; ++f) { size_t end = filevec[f].buffered_lines; char *discards = discarded[f]; GNULineRef *counts = equiv_count[1 - f]; GNULineRef *equivs = filevec[f].equivs; size_t many = 5; size_t tem = end / 64; /* Multiply MANY by approximate square root of number of lines. That is the threshold for provisionally discardable lines. */ while((tem = tem >> 2) > 0) many *= 2; for(i = 0; i < (GNULineRef)end; ++i) { GNULineRef nmatch; if(equivs[i] == 0) continue; nmatch = counts[equivs[i]]; if(nmatch == 0) discards[i] = 1; else if(nmatch > (GNULineRef)many) discards[i] = 2; } } /* Don't really discard the provisional lines except when they occur in a run of discardables, with nonprovisionals at the beginning and end. */ for(f = 0; f < 2; ++f) { GNULineRef end = filevec[f].buffered_lines; char *discards = discarded[f]; for(i = 0; i < end; ++i) { /* Cancel provisional discards not in middle of run of discards. */ if(discards[i] == 2) discards[i] = 0; else if(discards[i] != 0) { /* We have found a nonprovisional discard. */ GNULineRef j; GNULineRef length; GNULineRef provisional = 0; /* Find end of this run of discardable lines. Count how many are provisionally discardable. */ for(j = i; j < end; ++j) { if(discards[j] == 0) break; if(discards[j] == 2) ++provisional; } /* Cancel provisional discards at end, and shrink the run. */ while(j > i && discards[j - 1] == 2) discards[--j] = 0, --provisional; /* Now we have the length of a run of discardable lines whose first and last are not provisional. */ length = j - i; /* If 1/4 of the lines in the run are provisional, cancel discarding of all provisional lines in the run. */ if(provisional * 4 > length) { while(j > i) if(discards[--j] == 2) discards[j] = 0; } else { GNULineRef consec; GNULineRef minimum = 1; GNULineRef tem = length >> 2; /* MINIMUM is approximate square root of LENGTH/4. A subrun of two or more provisionals can stand when LENGTH is at least 16. A subrun of 4 or more can stand when LENGTH >= 64. */ while(0 < (tem >>= 2)) minimum <<= 1; minimum++; /* Cancel any subrun of MINIMUM or more provisionals within the larger run. */ for(j = 0, consec = 0; j < length; ++j) if(discards[i + j] != 2) consec = 0; else if(minimum == ++consec) /* Back up to start of subrun, to cancel it all. */ j -= consec; else if(minimum < consec) discards[i + j] = 0; /* Scan from beginning of run until we find 3 or more nonprovisionals in a row or until the first nonprovisional at least 8 lines in. Until that point, cancel any provisionals. */ for(j = 0, consec = 0; j < length; ++j) { if(j >= 8 && discards[i + j] == 1) break; if(discards[i + j] == 2) consec = 0, discards[i + j] = 0; else if(discards[i + j] == 0) consec = 0; else consec++; if(consec == 3) break; } /* I advances to the last line of the run. */ i += length - 1; /* Same thing, from end. */ for(j = 0, consec = 0; j < length; ++j) { if(j >= 8 && discards[i - j] == 1) break; if(discards[i - j] == 2) consec = 0, discards[i - j] = 0; else if(discards[i - j] == 0) consec = 0; else consec++; if(consec == 3) break; } } } } } /* Actually discard the lines. */ for(f = 0; f < 2; ++f) { char *discards = discarded[f]; GNULineRef end = filevec[f].buffered_lines; GNULineRef j = 0; for(i = 0; i < end; ++i) if(minimal || discards[i] == 0) { filevec[f].undiscarded[j] = filevec[f].equivs[i]; filevec[f].realindexes[j++] = i; } else filevec[f].changed[i] = true; filevec[f].nondiscarded_lines = j; } free(discarded[0]); free(equiv_count[0]); } /* Adjust inserts/deletes of identical lines to join changes as much as possible. We do something when a run of changed lines include a line at one end and have an excluded, identical line at the other. We are free to choose which identical line is included. `compareseq' usually chooses the one at the beginning, but usually it is cleaner to consider the following identical line to be the "change". */ void GnuDiff::shift_boundaries(file_data filevec[]) { int f; for(f = 0; f < 2; ++f) { bool *changed = filevec[f].changed; bool const *other_changed = filevec[1 - f].changed; GNULineRef const *equivs = filevec[f].equivs; GNULineRef i = 0; GNULineRef j = 0; GNULineRef i_end = filevec[f].buffered_lines; while(true) { GNULineRef runlength, start, corresponding; /* Scan forwards to find beginning of another run of changes. Also keep track of the corresponding point in the other file. */ while(i < i_end && !changed[i]) { while(other_changed[j++]) continue; i++; } if(i == i_end) break; start = i; /* Find the end of this run of changes. */ while(changed[++i]) continue; while(other_changed[j]) j++; do { /* Record the length of this run of changes, so that we can later determine whether the run has grown. */ runlength = i - start; /* Move the changed region back, so long as the previous unchanged line matches the last changed one. This merges with previous changed regions. */ while(start && equivs[start - 1] == equivs[i - 1]) { changed[--start] = true; changed[--i] = false; while(changed[start - 1]) start--; while(other_changed[--j]) continue; } /* Set CORRESPONDING to the end of the changed run, at the last point where it corresponds to a changed run in the other file. CORRESPONDING == I_END means no such point has been found. */ corresponding = other_changed[j - 1] ? i : i_end; /* Move the changed region forward, so long as the first changed line matches the following unchanged one. This merges with following changed regions. Do this second, so that if there are no merges, the changed region is moved forward as far as possible. */ while(i != i_end && equivs[start] == equivs[i]) { changed[start++] = false; changed[i++] = true; while(changed[i]) i++; while(other_changed[++j]) corresponding = i; } } while(runlength != i - start); /* If possible, move the fully-merged run of changes back to a corresponding run in the other file. */ while(corresponding < i) { changed[--start] = true; changed[--i] = false; while(other_changed[--j]) continue; } } } } /* Cons an additional entry onto the front of an edit script OLD. LINE0 and LINE1 are the first affected lines in the two files (origin 0). DELETED is the number of lines deleted here from file 0. INSERTED is the number of lines inserted here in file 1. If DELETED is 0 then LINE0 is the number of the line before which the insertion was done; vice versa for INSERTED and LINE1. */ GnuDiff::change *GnuDiff::add_change(GNULineRef line0, GNULineRef line1, GNULineRef deleted, GNULineRef inserted, change *old) { - change *newChange = (change *)xmalloc(sizeof *newChange); + change *newChange = (change *)xmalloc(sizeof(*newChange)); newChange->line0 = line0; newChange->line1 = line1; newChange->inserted = inserted; newChange->deleted = deleted; newChange->link = old; return newChange; } /* Scan the tables of which lines are inserted and deleted, producing an edit script in reverse order. */ GnuDiff::change *GnuDiff::build_reverse_script(file_data const filevec[]) { change *script = nullptr; bool *changed0 = filevec[0].changed; bool *changed1 = filevec[1].changed; GNULineRef len0 = filevec[0].buffered_lines; GNULineRef len1 = filevec[1].buffered_lines; /* Note that changedN[len0] does exist, and is 0. */ GNULineRef i0 = 0, i1 = 0; while(i0 < len0 || i1 < len1) { if(changed0[i0] | changed1[i1]) { GNULineRef line0 = i0, line1 = i1; /* Find # lines changed here in each file. */ while(changed0[i0]) ++i0; while(changed1[i1]) ++i1; /* Record this change. */ script = add_change(line0, line1, i0 - line0, i1 - line1, script); } /* We have reached lines in the two files that match each other. */ i0++, i1++; } return script; } /* Scan the tables of which lines are inserted and deleted, producing an edit script in forward order. */ GnuDiff::change *GnuDiff::build_script(file_data const filevec[]) { change *script = nullptr; bool *changed0 = filevec[0].changed; bool *changed1 = filevec[1].changed; GNULineRef i0 = filevec[0].buffered_lines, i1 = filevec[1].buffered_lines; /* Note that changedN[-1] does exist, and is 0. */ while(i0 >= 0 || i1 >= 0) { if(changed0[i0 - 1] | changed1[i1 - 1]) { GNULineRef line0 = i0, line1 = i1; /* Find # lines changed here in each file. */ while(changed0[i0 - 1]) --i0; while(changed1[i1 - 1]) --i1; /* Record this change. */ script = add_change(i0, i1, line0 - i0, line1 - i1, script); } /* We have reached lines in the two files that match each other. */ i0--, i1--; } return script; } /* Report the differences of two files. */ GnuDiff::change *GnuDiff::diff_2_files(comparison *cmp) { GNULineRef diags; int f; change *script; read_files(cmp->file, files_can_be_treated_as_binary); { /* Allocate vectors for the results of comparison: a flag for each line of each file, saying whether that line is an insertion or deletion. Allocate an extra element, always 0, at each end of each vector. */ size_t s = cmp->file[0].buffered_lines + cmp->file[1].buffered_lines + 4; bool *flag_space = (bool *)zalloc(s * sizeof(*flag_space)); cmp->file[0].changed = flag_space + 1; cmp->file[1].changed = flag_space + cmp->file[0].buffered_lines + 3; /* Some lines are obviously insertions or deletions because they don't match anything. Detect them now, and avoid even thinking about them in the main comparison algorithm. */ discard_confusing_lines(cmp->file); /* Now do the main comparison algorithm, considering just the undiscarded lines. */ xvec = cmp->file[0].undiscarded; yvec = cmp->file[1].undiscarded; diags = (cmp->file[0].nondiscarded_lines + cmp->file[1].nondiscarded_lines + 3); - fdiag = (GNULineRef *)xmalloc(diags * (2 * sizeof *fdiag)); + fdiag = (GNULineRef *)xmalloc(diags * (2 * sizeof(*fdiag))); bdiag = fdiag + diags; fdiag += cmp->file[1].nondiscarded_lines + 1; bdiag += cmp->file[1].nondiscarded_lines + 1; /* Set TOO_EXPENSIVE to be approximate square root of input size, bounded below by 256. */ too_expensive = 1; for(; diags != 0; diags >>= 2) too_expensive <<= 1; too_expensive = std::max((GNULineRef)256, too_expensive); files[0] = cmp->file[0]; files[1] = cmp->file[1]; compareseq(0, cmp->file[0].nondiscarded_lines, 0, cmp->file[1].nondiscarded_lines, minimal); free(fdiag - (cmp->file[1].nondiscarded_lines + 1)); /* Modify the results slightly to make them prettier in cases where that can validly be done. */ shift_boundaries(cmp->file); /* Get the results of comparison in the form of a chain of `change's -- an edit script. */ script = build_script(cmp->file); free(cmp->file[0].undiscarded); free(flag_space); for(f = 0; f < 2; ++f) { free(cmp->file[f].equivs); free(cmp->file[f].linbuf + cmp->file[f].linbuf_base); } } return script; } diff --git a/src/gnudiff_diff.h b/src/gnudiff_diff.h index 146f2b5..b07a74d 100644 --- a/src/gnudiff_diff.h +++ b/src/gnudiff_diff.h @@ -1,228 +1,228 @@ /* Shared definitions for GNU DIFF Modified for KDiff3 by Joachim Eibl 2003, 2004, 2005. The original file was part of GNU DIFF. Copyright (C) 1988, 1989, 1991, 1992, 1993, 1994, 1995, 1998, 2001, 2002 Free Software Foundation, Inc. GNU DIFF is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version. GNU DIFF is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; see the file COPYING. If not, write to the Free Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ #ifndef GNUDIFF_DIFF_H #define GNUDIFF_DIFF_H #include "LineRef.h" #include "Utils.h" #include #include #include #include #include #include #include #include #include #include #include /* The integer type of a line number. */ typedef qint64 GNULineRef; #define GNULINEREF_MAX std::numeric_limits::max() static_assert(std::is_signed::value, "GNULineRef must be signed."); static_assert(sizeof(GNULineRef) >= sizeof(size_t), "GNULineRef must be able to receive size_t values."); class GnuDiff { public: /* Variables for command line options */ /* Nonzero if output cannot be generated for identical files. */ bool no_diff_means_no_output; /* Number of lines of context to show in each set of diffs. This is zero when context is not to be shown. */ GNULineRef context; /* The significance of white space during comparisons. */ enum { /* All white space is significant (the default). */ IGNORE_NO_WHITE_SPACE, /* Ignore changes due to tab expansion (-E). */ IGNORE_TAB_EXPANSION, /* Ignore changes in horizontal white space (-b). */ IGNORE_SPACE_CHANGE, /* Ignore all horizontal white space (-w). */ IGNORE_ALL_SPACE } ignore_white_space; /* Ignore changes that affect only numbers. (J. Eibl) */ bool bIgnoreNumbers; bool bIgnoreWhiteSpace; /* Files can be compared byte-by-byte, as if they were binary. This depends on various options. */ bool files_can_be_treated_as_binary; /* Ignore differences in case of letters (-i). */ bool ignore_case; /* Use heuristics for better speed with large files with a small density of changes. */ bool speed_large_files; /* Don't discard lines. This makes things slower (sometimes much slower) but will find a guaranteed minimal set of changes. */ bool minimal; /* The result of comparison is an "edit script": a chain of `struct change'. Each `struct change' represents one place where some lines are deleted and some are inserted. LINE0 and LINE1 are the first affected lines in the two files (origin 0). DELETED is the number of lines deleted here from file 0. INSERTED is the number of lines inserted here in file 1. If DELETED is 0 then LINE0 is the number of the line before which the insertion was done; vice versa for INSERTED and LINE1. */ struct change { change *link; /* Previous or next edit command */ GNULineRef inserted; /* # lines of file 1 changed here. */ GNULineRef deleted; /* # lines of file 0 changed here. */ GNULineRef line0; /* Line number of 1st deleted line. */ GNULineRef line1; /* Line number of 1st inserted line. */ bool ignore; /* Flag used in context.c. */ }; /* Structures that describe the input files. */ /* Data on one input file being compared. */ struct file_data { /* Buffer in which text of file is read. */ const QChar *buffer; /* Allocated size of buffer, in QChars. Always a multiple of - sizeof *buffer. */ + sizeof(*buffer). */ size_t bufsize; /* Number of valid bytes now in the buffer. */ size_t buffered; /* Array of pointers to lines in the file. */ const QChar **linbuf; /* linbuf_base <= buffered_lines <= valid_lines <= alloc_lines. linebuf[linbuf_base ... buffered_lines - 1] are possibly differing. linebuf[linbuf_base ... valid_lines - 1] contain valid data. linebuf[linbuf_base ... alloc_lines - 1] are allocated. */ GNULineRef linbuf_base, buffered_lines, valid_lines, alloc_lines; /* Pointer to end of prefix of this file to ignore when hashing. */ const QChar *prefix_end; /* Count of lines in the prefix. There are this many lines in the file before linbuf[0]. */ GNULineRef prefix_lines; /* Pointer to start of suffix of this file to ignore when hashing. */ const QChar *suffix_begin; /* Vector, indexed by line number, containing an equivalence code for each line. It is this vector that is actually compared with that of another file to generate differences. */ GNULineRef *equivs; /* Vector, like the previous one except that the elements for discarded lines have been squeezed out. */ GNULineRef *undiscarded; /* Vector mapping virtual line numbers (not counting discarded lines) to real ones (counting those lines). Both are origin-0. */ GNULineRef *realindexes; /* Total number of nondiscarded lines. */ GNULineRef nondiscarded_lines; /* Vector, indexed by real origin-0 line number, containing TRUE for a line that is an insertion or a deletion. The results of comparison are stored here. */ bool *changed; /* 1 if at end of file. */ bool eof; /* 1 more than the maximum equivalence value used for this or its sibling file. */ GNULineRef equiv_max; }; /* Data on two input files being compared. */ struct comparison { file_data file[2]; comparison const *parent; /* parent, if a recursive comparison */ }; /* Describe the two files currently being compared. */ file_data files[2]; /* Declare various functions. */ /* analyze.c */ change *diff_2_files(comparison *); /* io.c */ bool read_files(file_data[], bool); /* util.c */ bool lines_differ(const QChar *, size_t, const QChar *, size_t); void *zalloc(size_t); private: // gnudiff_analyze.cpp GNULineRef diag(GNULineRef xoff, GNULineRef xlim, GNULineRef yoff, GNULineRef ylim, bool find_minimal, struct partition *part); void compareseq(GNULineRef xoff, GNULineRef xlim, GNULineRef yoff, GNULineRef ylim, bool find_minimal); void discard_confusing_lines(file_data filevec[]); void shift_boundaries(file_data filevec[]); change *add_change(GNULineRef line0, GNULineRef line1, GNULineRef deleted, GNULineRef inserted, change *old); change *build_reverse_script(file_data const filevec[]); change *build_script(file_data const filevec[]); // gnudiff_io.cpp void find_and_hash_each_line(file_data *current); void find_identical_ends(file_data filevec[]); // gnudiff_xmalloc.cpp void *xmalloc(size_t n); void *xrealloc(void *p, size_t n); void xalloc_die(); inline bool isWhite(QChar c) { return c == ' ' || c == '\t' || c == '\r'; } }; // class GnuDiff #endif diff --git a/src/gnudiff_io.cpp b/src/gnudiff_io.cpp index a0e919d..639c877 100644 --- a/src/gnudiff_io.cpp +++ b/src/gnudiff_io.cpp @@ -1,545 +1,545 @@ /* File I/O for GNU DIFF. Modified for KDiff3 by Joachim Eibl 2003, 2004, 2005. The original file was part of GNU DIFF. Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1998, 2001, 2002 Free Software Foundation, Inc. GNU DIFF is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version. GNU DIFF is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; see the file COPYING. If not, write to the Free Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ #include "gnudiff_diff.h" #include #include /* Rotate an unsigned value to the left. */ #define ROL(v, n) ((v) << (n) | (v) >> (sizeof(v) * CHAR_BIT - (n))) /* Given a hash value and a new character, return a new hash value. */ #define HASH(h, c) ((c) + ROL(h, 7)) /* The type of a hash value. */ typedef size_t hash_value; static_assert(std::is_unsigned::value, "hash_value must be signed."); /* Lines are put into equivalence classes of lines that match in lines_differ. Each equivalence class is represented by one of these structures, but only while the classes are being computed. Afterward, each class is represented by a number. */ struct equivclass { GNULineRef next; /* Next item in this bucket. */ hash_value hash; /* Hash of lines in this class. */ const QChar *line; /* A line that fits this class. */ size_t length; /* That line's length, not counting its newline. */ }; /* Hash-table: array of buckets, each being a chain of equivalence classes. buckets[-1] is reserved for incomplete lines. */ static GNULineRef *buckets; /* Number of buckets in the hash table array, not counting buckets[-1]. */ static size_t nbuckets; /* Array in which the equivalence classes are allocated. The bucket-chains go through the elements in this array. The number of an equivalence class is its index in this array. */ static equivclass *equivs; /* Index of first free element in the array `equivs'. */ static GNULineRef equivs_index; /* Number of elements allocated in the array `equivs'. */ static GNULineRef equivs_alloc; /* Check for binary files and compare them for exact identity. */ /* Return 1 if BUF contains a non text character. SIZE is the number of characters in BUF. */ #define binary_file_p(buf, size) (memchr(buf, 0, size) != 0) /* Compare two lines (typically one from each input file) according to the command line options. For efficiency, this is invoked only when the lines do not match exactly but an option like -i might cause us to ignore the difference. Return nonzero if the lines differ. */ bool GnuDiff::lines_differ(const QChar *s1, size_t len1, const QChar *s2, size_t len2) { const QChar *t1 = s1; const QChar *t2 = s2; const QChar *s1end = s1 + len1; const QChar *s2end = s2 + len2; for(;; ++t1, ++t2) { /* Test for exact char equality first, since it's a common case. */ if(t1 != s1end && t2 != s2end && *t1 == *t2) continue; else { while(t1 != s1end && ((bIgnoreWhiteSpace && isWhite(*t1)) || (bIgnoreNumbers && (t1->isDigit() || *t1 == '-' || *t1 == '.')))) { ++t1; } while(t2 != s2end && ((bIgnoreWhiteSpace && isWhite(*t2)) || (bIgnoreNumbers && (t2->isDigit() || *t2 == '-' || *t2 == '.')))) { ++t2; } if(t1 != s1end && t2 != s2end) { if(ignore_case) { /* Lowercase comparison. */ if(t1->toLower() == t2->toLower()) continue; } else if(*t1 == *t2) continue; else return true; } else if(t1 == s1end && t2 == s2end) return false; else return true; } } return false; } /* Split the file into lines, simultaneously computing the equivalence class for each line. */ void GnuDiff::find_and_hash_each_line(file_data *current) { hash_value h; const QChar *p = current->prefix_end; QChar c; GNULineRef i, *bucket; size_t length; /* Cache often-used quantities in local variables to help the compiler. */ const QChar **linbuf = current->linbuf; GNULineRef alloc_lines = current->alloc_lines; GNULineRef line = 0; GNULineRef linbuf_base = current->linbuf_base; - GNULineRef *cureqs = (GNULineRef *)xmalloc(alloc_lines * sizeof *cureqs); + GNULineRef *cureqs = (GNULineRef *)xmalloc(alloc_lines * sizeof(*cureqs)); equivclass *eqs = equivs; GNULineRef eqs_index = equivs_index; GNULineRef eqs_alloc = equivs_alloc; const QChar *suffix_begin = current->suffix_begin; const QChar *bufend = current->buffer + current->buffered; bool diff_length_compare_anyway = ignore_white_space != IGNORE_NO_WHITE_SPACE || bIgnoreNumbers; bool same_length_diff_contents_compare_anyway = diff_length_compare_anyway | ignore_case; while(p < suffix_begin) { const QChar *ip = p; h = 0; /* Hash this line until we find a newline or bufend is reached. */ if(ignore_case) switch(ignore_white_space) { case IGNORE_ALL_SPACE: while(p < bufend && !Utils::isEndOfLine(c = *p)) { if(!(isWhite(c) || (bIgnoreNumbers && (c.isDigit() || c == '-' || c == '.')))) h = HASH(h, c.toLower().unicode()); ++p; } break; default: while(p < bufend && !Utils::isEndOfLine(c = *p)) { h = HASH(h, c.toLower().unicode()); ++p; } break; } else switch(ignore_white_space) { case IGNORE_ALL_SPACE: while(p < bufend && !Utils::isEndOfLine(c = *p)) { if(!(isWhite(c) || (bIgnoreNumbers && (c.isDigit() || c == '-' || c == '.')))) h = HASH(h, c.unicode()); ++p; } break; default: while(p < bufend && !Utils::isEndOfLine(c = *p)) { h = HASH(h, c.unicode()); ++p; } break; } bucket = &buckets[h % nbuckets]; length = p - ip; ++p; for(i = *bucket;; i = eqs[i].next) if(!i) { /* Create a new equivalence class in this bucket. */ i = eqs_index++; if(i == eqs_alloc) { - if((GNULineRef)(GNULINEREF_MAX / (2 * sizeof *eqs)) <= eqs_alloc) + if((GNULineRef)(GNULINEREF_MAX / (2 * sizeof(*eqs))) <= eqs_alloc) xalloc_die(); eqs_alloc *= 2; - eqs = (equivclass *)xrealloc(eqs, eqs_alloc * sizeof *eqs); + eqs = (equivclass *)xrealloc(eqs, eqs_alloc * sizeof(*eqs)); } eqs[i].next = *bucket; eqs[i].hash = h; eqs[i].line = ip; eqs[i].length = length; *bucket = i; break; } else if(eqs[i].hash == h) { const QChar *eqline = eqs[i].line; /* Reuse existing class if lines_differ reports the lines equal. */ if(eqs[i].length == length) { /* Reuse existing equivalence class if the lines are identical. This detects the common case of exact identity faster than lines_differ would. */ if(memcmp(eqline, ip, length * sizeof(QChar)) == 0) break; if(!same_length_diff_contents_compare_anyway) continue; } else if(!diff_length_compare_anyway) continue; if(!lines_differ(eqline, eqs[i].length, ip, length)) break; } /* Maybe increase the size of the line table. */ if(line == alloc_lines) { /* Double (alloc_lines - linbuf_base) by adding to alloc_lines. */ - if((GNULineRef)(GNULINEREF_MAX / 3) <= alloc_lines || (GNULineRef)(GNULINEREF_MAX / sizeof *cureqs) <= 2 * alloc_lines - linbuf_base || (GNULineRef)(GNULINEREF_MAX / sizeof *linbuf) <= alloc_lines - linbuf_base) + if((GNULineRef)(GNULINEREF_MAX / 3) <= alloc_lines || (GNULineRef)(GNULINEREF_MAX / sizeof(*cureqs)) <= 2 * alloc_lines - linbuf_base || (GNULineRef)(GNULINEREF_MAX / sizeof(*linbuf)) <= alloc_lines - linbuf_base) xalloc_die(); alloc_lines = 2 * alloc_lines - linbuf_base; - cureqs = (GNULineRef *)xrealloc(cureqs, alloc_lines * sizeof *cureqs); + cureqs = (GNULineRef *)xrealloc(cureqs, alloc_lines * sizeof(*cureqs)); linbuf += linbuf_base; linbuf = (const QChar **)xrealloc(linbuf, - (alloc_lines - linbuf_base) * sizeof *linbuf); + (alloc_lines - linbuf_base) * sizeof(*linbuf)); linbuf -= linbuf_base; } linbuf[line] = ip; cureqs[line] = i; ++line; } current->buffered_lines = line; for(i = 0;; ++i) { /* Record the line start for lines in the suffix that we care about. Record one more line start than lines, so that we can compute the length of any buffered line. */ if(line == alloc_lines) { /* Double (alloc_lines - linbuf_base) by adding to alloc_lines. */ - if((GNULineRef)(GNULINEREF_MAX / 3) <= alloc_lines || (GNULineRef)(GNULINEREF_MAX / sizeof *cureqs) <= 2 * alloc_lines - linbuf_base || (GNULineRef)(GNULINEREF_MAX / sizeof *linbuf) <= alloc_lines - linbuf_base) + if((GNULineRef)(GNULINEREF_MAX / 3) <= alloc_lines || (GNULineRef)(GNULINEREF_MAX / sizeof(*cureqs)) <= 2 * alloc_lines - linbuf_base || (GNULineRef)(GNULINEREF_MAX / sizeof(*linbuf)) <= alloc_lines - linbuf_base) xalloc_die(); alloc_lines = 2 * alloc_lines - linbuf_base; linbuf += linbuf_base; linbuf = (const QChar **)xrealloc(linbuf, - (alloc_lines - linbuf_base) * sizeof *linbuf); + (alloc_lines - linbuf_base) * sizeof(*linbuf)); linbuf -= linbuf_base; } linbuf[line] = p; if(p >= bufend) break; if(context <= i && no_diff_means_no_output) break; line++; while(p < bufend && !Utils::isEndOfLine(*p++)) continue; } /* Done with cache in local variables. */ current->linbuf = linbuf; current->valid_lines = line; current->alloc_lines = alloc_lines; current->equivs = cureqs; equivs = eqs; equivs_alloc = eqs_alloc; equivs_index = eqs_index; } /* We have found N lines in a buffer of size S; guess the proportionate number of lines that will be found in a buffer of size T. However, do not guess a number of lines so large that the resulting line table might cause overflow in size calculations. */ static GNULineRef guess_lines(GNULineRef n, size_t s, size_t t) { size_t guessed_bytes_per_line = n < 10 ? 32 : s / (n - 1); size_t guessed_lines = std::max((size_t)1, t / guessed_bytes_per_line); return (GNULineRef)std::min((GNULineRef)guessed_lines, (GNULineRef)(GNULINEREF_MAX / (2 * sizeof(QChar *) + 1) - 5)) + 5; } /* Given a vector of two file_data objects, find the identical prefixes and suffixes of each object. */ void GnuDiff::find_identical_ends(file_data filevec[]) { /* Find identical prefix. */ const QChar *p0, *p1, *buffer0, *buffer1; p0 = buffer0 = filevec[0].buffer; p1 = buffer1 = filevec[1].buffer; size_t n0, n1; n0 = filevec[0].buffered; n1 = filevec[1].buffered; const QChar *const pEnd0 = p0 + n0; const QChar *const pEnd1 = p1 + n1; if(p0 == p1) /* The buffers are the same; sentinels won't work. */ p0 = p1 += n1; else { /* Loop until first mismatch, or end. */ while(p0 != pEnd0 && p1 != pEnd1 && *p0 == *p1) { p0++; p1++; } } /* Now P0 and P1 point at the first nonmatching characters. */ /* Skip back to last line-beginning in the prefix. */ while(p0 != buffer0 && !Utils::isEndOfLine(p0[-1])) p0--, p1--; /* Record the prefix. */ filevec[0].prefix_end = p0; filevec[1].prefix_end = p1; /* Find identical suffix. */ /* P0 and P1 point beyond the last chars not yet compared. */ p0 = buffer0 + n0; p1 = buffer1 + n1; const QChar *end0, *beg0; end0 = p0; /* Addr of last char in file 0. */ /* Get value of P0 at which we should stop scanning backward: this is when either P0 or P1 points just past the last char of the identical prefix. */ beg0 = filevec[0].prefix_end + (n0 < n1 ? 0 : n0 - n1); /* Scan back until chars don't match or we reach that point. */ for(; p0 != beg0; p0--, p1--) { if(*p0 != *p1) { /* Point at the first char of the matching suffix. */ beg0 = p0; break; } } // Go to the next line (skip last line with a difference) if(p0 != end0) { if(*p0 != *p1) ++p0; while(p0 < pEnd0 && !Utils::isEndOfLine(*p0++)) continue; } p1 += p0 - beg0; /* Record the suffix. */ filevec[0].suffix_begin = p0; filevec[1].suffix_begin = p1; /* Calculate number of lines of prefix to save. prefix_count == 0 means save the whole prefix; we need this for options like -D that output the whole file, or for enormous contexts (to avoid worrying about arithmetic overflow). We also need it for options like -F that output some preceding line; at least we will need to find the last few lines, but since we don't know how many, it's easiest to find them all. Otherwise, prefix_count != 0. Save just prefix_count lines at start of the line buffer; they'll be moved to the proper location later. Handle 1 more line than the context says (because we count 1 too many), rounded up to the next power of 2 to speed index computation. */ const QChar **linbuf0, **linbuf1; GNULineRef alloc_lines0, alloc_lines1; GNULineRef buffered_prefix, prefix_count, prefix_mask; GNULineRef middle_guess, suffix_guess; if(no_diff_means_no_output && context < (GNULineRef)(GNULINEREF_MAX / 4) && context < (GNULineRef)(n0)) { middle_guess = guess_lines(0, 0, p0 - filevec[0].prefix_end); suffix_guess = guess_lines(0, 0, buffer0 + n0 - p0); for(prefix_count = 1; prefix_count <= context; prefix_count *= 2) continue; alloc_lines0 = (prefix_count + middle_guess + std::min(context, suffix_guess)); } else { prefix_count = 0; alloc_lines0 = guess_lines(0, 0, n0); } prefix_mask = prefix_count - 1; GNULineRef lines = 0; linbuf0 = (const QChar **)xmalloc(alloc_lines0 * sizeof(*linbuf0)); p0 = buffer0; /* If the prefix is needed, find the prefix lines. */ if(!(no_diff_means_no_output && filevec[0].prefix_end == p0 && filevec[1].prefix_end == p1)) { end0 = filevec[0].prefix_end; while(p0 != end0) { GNULineRef l = lines++ & prefix_mask; if(l == alloc_lines0) { - if((GNULineRef)(GNULINEREF_MAX / (2 * sizeof *linbuf0)) <= alloc_lines0) + if((GNULineRef)(GNULINEREF_MAX / (2 * sizeof(*linbuf0))) <= alloc_lines0) xalloc_die(); alloc_lines0 *= 2; linbuf0 = (const QChar **)xrealloc(linbuf0, alloc_lines0 * sizeof(*linbuf0)); } linbuf0[l] = p0; while(p0 < pEnd0 && !Utils::isEndOfLine(*p0++)) continue; } } buffered_prefix = prefix_count && context < lines ? context : lines; /* Allocate line buffer 1. */ middle_guess = guess_lines(lines, p0 - buffer0, p1 - filevec[1].prefix_end); suffix_guess = guess_lines(lines, p0 - buffer0, buffer1 + n1 - p1); alloc_lines1 = buffered_prefix + middle_guess + std::min(context, suffix_guess); - if(alloc_lines1 < buffered_prefix || (GNULineRef)(GNULINEREF_MAX / sizeof *linbuf1) <= alloc_lines1) + if(alloc_lines1 < buffered_prefix || (GNULineRef)(GNULINEREF_MAX / sizeof(*linbuf1)) <= alloc_lines1) xalloc_die(); linbuf1 = (const QChar **)xmalloc(alloc_lines1 * sizeof(*linbuf1)); GNULineRef i; if(buffered_prefix != lines) { /* Rotate prefix lines to proper location. */ for(i = 0; i < buffered_prefix; ++i) linbuf1[i] = linbuf0[(lines - context + i) & prefix_mask]; for(i = 0; i < buffered_prefix; ++i) linbuf0[i] = linbuf1[i]; } /* Initialize line buffer 1 from line buffer 0. */ for(i = 0; i < buffered_prefix; ++i) linbuf1[i] = linbuf0[i] - buffer0 + buffer1; /* Record the line buffer, adjusted so that linbuf[0] points at the first differing line. */ filevec[0].linbuf = linbuf0 + buffered_prefix; filevec[1].linbuf = linbuf1 + buffered_prefix; filevec[0].linbuf_base = filevec[1].linbuf_base = -buffered_prefix; filevec[0].alloc_lines = alloc_lines0 - buffered_prefix; filevec[1].alloc_lines = alloc_lines1 - buffered_prefix; filevec[0].prefix_lines = filevec[1].prefix_lines = lines; } /* If 1 < k, then (2**k - prime_offset[k]) is the largest prime less than 2**k. This table is derived from Chris K. Caldwell's list . */ static unsigned char const prime_offset[] = { 0, 0, 1, 1, 3, 1, 3, 1, 5, 3, 3, 9, 3, 1, 3, 19, 15, 1, 5, 1, 3, 9, 3, 15, 3, 39, 5, 39, 57, 3, 35, 1, 5, 9, 41, 31, 5, 25, 45, 7, 87, 21, 11, 57, 17, 55, 21, 115, 59, 81, 27, 129, 47, 111, 33, 55, 5, 13, 27, 55, 93, 1, 57, 25}; /* Verify that this host's size_t is not too wide for the above table. */ static_assert(sizeof(size_t) * CHAR_BIT <= sizeof prime_offset, "Not enough primes in table"); /* Given a vector of two file_data objects, read the file associated with each one, and build the table of equivalence classes. Return nonzero if either file appears to be a binary file. If PRETEND_BINARY is nonzero, pretend they are binary regardless. */ bool GnuDiff::read_files(file_data filevec[], bool /*pretend_binary*/) { GNULineRef i; find_identical_ends(filevec); equivs_alloc = filevec[0].alloc_lines + filevec[1].alloc_lines + 1; - if((GNULineRef)(GNULINEREF_MAX / sizeof *equivs) <= equivs_alloc) + if((GNULineRef)(GNULINEREF_MAX / sizeof(*equivs)) <= equivs_alloc) xalloc_die(); - equivs = (equivclass *)xmalloc(equivs_alloc * sizeof *equivs); + equivs = (equivclass *)xmalloc(equivs_alloc * sizeof(*equivs)); /* Equivalence class 0 is permanently safe for lines that were not hashed. Real equivalence classes start at 1. */ equivs_index = 1; /* Allocate (one plus) a prime number of hash buckets. Use a prime number between 1/3 and 2/3 of the value of equiv_allocs, approximately. */ for(i = 9; ((GNULineRef)1 << i) < equivs_alloc / 3; ++i) continue; nbuckets = ((GNULineRef)1 << i) - prime_offset[i]; - if(GNULINEREF_MAX / sizeof *buckets <= nbuckets) + if(GNULINEREF_MAX / sizeof(*buckets) <= nbuckets) xalloc_die(); - buckets = (GNULineRef *)zalloc((nbuckets + 1) * sizeof *buckets); + buckets = (GNULineRef *)zalloc((nbuckets + 1) * sizeof(*buckets)); buckets++; for(i = 0; i < 2; ++i) find_and_hash_each_line(&filevec[i]); filevec[0].equiv_max = filevec[1].equiv_max = equivs_index; free(equivs); free(buckets - 1); return false; }