lyx_mirror/src/lengthcommon.cpp
Jean-Marc Lasgouttes d9c9c6b36d Remove C-isms in isValidLength
This fixes coverity issue 23349, with a potential out-of-bounds write to the pattern array.
2015-02-20 16:16:31 +01:00

368 lines
8.9 KiB
C++

/**
* \file lengthcommon.cpp
* This file is part of LyX, the document processor.
* Licence details can be found in the file COPYING.
*
* \author Lars Gullik Bjønnes
* \author Matthias Ettrich
* \author John Levon
*
* Full author contact details are available in file CREDITS.
*/
#include <config.h>
#include "Length.h"
#include "support/convert.h"
#include "support/gettext.h"
#include "support/lassert.h"
#include "support/lstrings.h"
using namespace std;
using namespace lyx::support;
namespace lyx {
// I am not sure if "mu" should be possible to select (Lgb)
// the latex units
char const * const unit_name[] = {
"bp", "cc", "cm", "dd", "em", "ex", "in", "mm", "mu",
"pc", "pt", "sp",
"text%", "col%", "page%", "line%",
"theight%", "pheight%", "" };
int const num_units = int(sizeof(unit_name) / sizeof(unit_name[0]) - 1);
// the LyX gui units
char const * const unit_name_gui[] = {
N_("bp"), N_("cc[[unit of measure]]"), N_("cm"), N_("dd"), N_("em"),
N_("ex"), N_("in[[unit of measure]]"), N_("mm"), N_("mu[[unit of measure]]"), N_("pc"),
N_("pt"), N_("sp"), N_("Text Width %"),
N_("Column Width %"), N_("Page Width %"), N_("Line Width %"),
N_("Text Height %"), N_("Page Height %"), "" };
Length::UNIT unitFromString(string const & data)
{
int i = 0;
while (i < num_units && data != unit_name[i])
++i;
return static_cast<Length::UNIT>(i);
}
namespace {
/// skip n characters of input
inline void lyx_advance(string & data, size_t n)
{
data.erase(0, n);
}
/// return true when the input is at the end
inline bool isEndOfData(string const & data)
{
return ltrim(data).empty();
}
/**
* nextToken - return the next token in the input
* @param data input string
* @param number_index the current position in the number array
* @param unit_index the current position in the unit array
* @return a char representing the type of token returned
*
* The possible return values are :
* + stretch indicator for glue length
* - shrink indicator for glue length
* n a numeric value (stored in number array)
* u a unit type (stored in unit array)
* E parse error
*/
char nextToken(string & data, double * number, int & number_index,
Length::UNIT * unit, int & unit_index)
{
data = ltrim(data);
if (data.empty())
return '\0';
if (data[0] == '+') {
lyx_advance(data, 1);
return '+';
}
if (prefixIs(data, "plus")) {
lyx_advance(data, 4);
return '+';
}
if (data[0] == '-') {
lyx_advance(data, 1);
return '-';
}
if (prefixIs(data, "minus")) {
lyx_advance(data, 5);
return '-';
}
size_t i = data.find_first_not_of("0123456789.");
if (i != 0) {
if (number_index > 3)
return 'E';
string buffer;
// we have found some number
if (i == string::npos) {
buffer = data;
i = data.size() + 1;
} else {
buffer = data.substr(0, i);
}
lyx_advance(data, i);
if (isStrDbl(buffer)) {
number[number_index] = convert<double>(buffer);
++number_index;
return 'n';
}
return 'E';
}
i = data.find_first_not_of("abcdefghijklmnopqrstuvwxyz%");
if (i != 0) {
if (unit_index > 3)
return 'E';
string buffer;
// we have found some alphabetical string
if (i == string::npos) {
buffer = data;
i = data.size() + 1;
} else {
buffer = data.substr(0, i);
}
// possibly we have "mmplus" string or similar
if (buffer.size() > 5 &&
(buffer.substr(2, 4) == string("plus") ||
buffer.substr(2, 5) == string("minus")))
{
lyx_advance(data, 2);
unit[unit_index] = unitFromString(buffer.substr(0, 2));
} else {
lyx_advance(data, i);
unit[unit_index] = unitFromString(buffer);
}
if (unit[unit_index] != Length::UNIT_NONE) {
++unit_index;
return 'u';
}
return 'E'; // Error
}
return 'E'; // Error
}
/// latex representation of a vspace
struct LaTeXLength {
char const * pattern;
int plus_val_index;
int minus_val_index;
int plus_uni_index;
int minus_uni_index;
};
/// the possible formats for a vspace string
LaTeXLength table[] = {
{ "nu", 0, 0, 0, 0 },
{ "nu+nu", 2, 0, 2, 0 },
{ "nu+nu-nu", 2, 3, 2, 3 },
{ "nu+-nu", 2, 2, 2, 2 },
{ "nu-nu", 0, 2, 0, 2 },
{ "nu-nu+nu", 3, 2, 3, 2 },
{ "nu-+nu", 2, 2, 2, 2 },
{ "n+nu", 2, 0, 1, 0 },
{ "n+n-nu", 2, 3, 1, 1 },
{ "n+-nu", 2, 2, 1, 1 },
{ "n-nu", 0, 2, 0, 1 },
{ "n-n+nu", 3, 2, 1, 1 },
{ "n-+nu", 2, 2, 1, 1 },
{ "", 0, 0, 0, 0 } // sentinel, must be empty
};
} // namespace anon
const char * stringFromUnit(int unit)
{
if (unit < 0 || unit > num_units)
return 0;
return unit_name[unit];
}
bool isValidGlueLength(string const & data, GlueLength * result)
{
// This parser is table-driven. First, it constructs a "pattern"
// that describes the sequence of tokens in "data". For example,
// "n-nu" means: number, minus sign, number, unit. As we go along,
// numbers and units are stored into static arrays. Then, "pattern"
// is searched in the "table". If it is found, the associated
// table entries tell us which number and unit should go where
// in the Length structure. Example: if "data" has the "pattern"
// "nu+nu-nu", the associated table entries are "2, 3, 2, 3".
// That means, "plus_val" is the second number that was seen
// in the input, "minus_val" is the third number, and "plus_uni"
// and "minus_uni" are the second and third units, respectively.
// ("val" and "uni" are always the first items seen in "data".)
// This is the most elegant solution I could find -- a straight-
// forward approach leads to very long, tedious code that would be
// much harder to understand and maintain. (AS)
if (data.empty()) {
if (result)
*result = GlueLength();
return true;
}
string buffer = ltrim(data);
// To make isValidGlueLength recognize negative values as
// the first number this little hack is needed:
int val_sign = 1; // positive as default
switch (buffer[0]) {
case '-':
lyx_advance(buffer, 1);
val_sign = -1;
break;
case '+':
lyx_advance(buffer, 1);
break;
default:
break;
}
// end of hack
// used to return numeric values in parsing vspace
double number[4] = { 0, 0, 0, 0 };
// used to return unit types in parsing vspace
Length::UNIT unit[4] = {Length::UNIT_NONE, Length::UNIT_NONE,
Length::UNIT_NONE, Length::UNIT_NONE};
int number_index = 1; // entries at index 0 are sentinels
int unit_index = 1; // entries at index 0 are sentinels
// construct "pattern" from "data"
size_t const pattern_max_size = 20;
string pattern;
while (!isEndOfData(buffer)) {
if (pattern.size() > pattern_max_size)
return false;
char const c = nextToken(buffer, number, number_index, unit,
unit_index);
if (c == 'E')
return false;
pattern.push_back(c);
}
// search "pattern" in "table"
size_t table_index = 0;
while (pattern != table[table_index].pattern) {
++table_index;
if (!*table[table_index].pattern)
return false;
}
// Get the values from the appropriate places. If an index
// is zero, the corresponding array value is zero or UNIT_NONE,
// so we needn't check this.
if (result) {
result->len_.value (number[1] * val_sign);
result->len_.unit (unit[1]);
result->plus_.value (number[table[table_index].plus_val_index]);
result->plus_.unit (unit [table[table_index].plus_uni_index]);
result->minus_.value(number[table[table_index].minus_val_index]);
result->minus_.unit (unit [table[table_index].minus_uni_index]);
}
return true;
}
bool isValidLength(string const & data, Length * result)
{
// This is a trimmed down version of isValidGlueLength.
// The parser may seem overkill for lengths without
// glue, but since we already have it, using it is
// easier than writing something from scratch.
if (data.empty()) {
if (result)
*result = Length();
return true;
}
string buffer = data;
// To make isValidLength recognize negative values
// this little hack is needed:
int val_sign = 1; // positive as default
switch (buffer[0]) {
case '-':
lyx_advance(buffer, 1);
val_sign = -1;
break;
case '+':
lyx_advance(buffer, 1);
// fall through
default:
// no action
break;
}
// end of hack
// used to return numeric values in parsing vspace
double number[4] = { 0, 0, 0, 0 };
// used to return unit types in parsing vspace
Length::UNIT unit[4] = {Length::UNIT_NONE, Length::UNIT_NONE,
Length::UNIT_NONE, Length::UNIT_NONE};
int number_index = 1; // entries at index 0 are sentinels
int unit_index = 1; // entries at index 0 are sentinels
// construct "pattern" from "data"
string pattern;
while (!isEndOfData(buffer)) {
if (pattern.size() > 2)
return false;
char const token = nextToken(buffer, number,
number_index, unit, unit_index);
if (token == 'E')
return false;
pattern += token;
}
// only the most basic pattern is accepted here
if (pattern != "nu")
return false;
// It _was_ a correct length string.
// Store away the values we found.
if (result) {
result->val_ = number[1] * val_sign;
result->unit_ = unit[1];
}
return true;
}
} // namespace lyx