Package: apt
Version: 0.7.12
Severity: normal
The apt configuration parser has an arbitrary limit of 1024 bytes
per line. This has recently bitten aptitude; I made some interface
strings configurable, and it turns out that in other languages such as
Russian these strings are more than 1024 bytes long.
I've come up with the least invasive patch I felt comfortable with
to lift the 1024-byte restriction, by doing the following:
(a) replacing the static array Buffer with a string named "Fragment"
that's built from repeated calls to getline().
(b) instead of iterating over the string and deleting chunks at the
beginning and end, keep beginning and end iterators and move
them around; to delete chunks from the middle, append to a new
string.
I've attached a patch, which is also pushed to the
improve-configuration-parsing branch.
These changes will make the configuration parser slightly slower and
slightly more memory-hungry. However, I would be shocked if they
mattered in practice: either version of the configuration parser should
be plenty fast on anything that will run apt. The startup time for
aptitude (measured with "aptitude nop") is essentially identical on my
system with either apt version.
In my own tests, my changes work fine and produce the same results
on files that used to work (e.g., from "apt-config dump"), but this is a
more invasive change than my last patch. I'd appreciate it if someone
else could review it.
Thanks,
Daniel
-- Package-specific info:
-- (/etc/apt/preferences present, but not submitted) --
-- (/etc/apt/sources.list present, but not submitted) --
-- System Information:
Debian Release: lenny/sid
Architecture: i386 (i686)
Kernel: Linux 2.6.24-1-686 (SMP w/1 CPU core)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8) (ignored: LC_ALL
set to en_US.UTF-8)
Shell: /bin/sh linked to /bin/bash
Versions of packages apt depends on:
ii debian-archive-keyring 2007.07.31 GnuPG archive keys of the Debian a
ii libc6 2.7-10 GNU C Library: Shared libraries
ii libgcc1 1:4.3.0-2 GCC support library
ii libstdc++6 4.3.0-2 The GNU Standard C++ Library v3
apt recommends no packages.
-- no debconf information
=== modified file 'apt-pkg/contrib/configuration.cc'
--- apt-pkg/contrib/configuration.cc 2008-04-01 04:54:01 +0000
+++ apt-pkg/contrib/configuration.cc 2008-04-02 03:55:05 +0000
@@ -495,8 +495,7 @@
ifstream F(FName.c_str(),ios::in);
if (!F != 0)
return _error->Errno("ifstream::ifstream",_("Opening configuration file %s"),FName.c_str());
-
- char Buffer[1024];
+
string LineBuffer;
string Stack[100];
unsigned int StackPos = 0;
@@ -508,24 +507,63 @@
bool InComment = false;
while (F.eof() == false)
{
- F.getline(Buffer,sizeof(Buffer));
+ // The raw input line.
+ std::string Input;
+ // The input line with comments stripped.
+ std::string Fragment;
+
+ // Grab the next line of F and place it in Input.
+ do
+ {
+ char *Buffer = new char[1024];
+
+ F.clear();
+ F.getline(Buffer,sizeof(Buffer) / 2);
+
+ Input += Buffer;
+ }
+ while (F.fail() && !F.eof());
+
+ // Expand tabs in the input line and remove leading and trailing
+ // whitespace.
+ {
+ const int BufferSize = Input.size() * 8 + 1;
+ char *Buffer = new char[BufferSize];
+ try
+ {
+ memcpy(Buffer, Input.c_str(), Input.size() + 1);
+
+ _strtabexpand(Buffer, BufferSize);
+ _strstrip(Buffer);
+ Input = Buffer;
+ }
+ catch(...)
+ {
+ delete[] Buffer;
+ throw;
+ }
+ delete[] Buffer;
+ }
CurLine++;
- // This should be made to work instead, but this is better than looping
- if (F.fail() && !F.eof())
- return _error->Error(_("%s: Line %d too long (max %lu)"),
- FName.c_str(), CurLine, sizeof(Buffer));
-
- _strtabexpand(Buffer,sizeof(Buffer));
- _strstrip(Buffer);
+
+ // Now strip comments; if the whole line is contained in a
+ // comment, skip this line.
+
+ // The first meaningful character in the current fragment; will
+ // be adjusted below as we remove bytes from the front.
+ std::string::const_iterator Start = Input.begin();
+ // The last meaningful character in the current fragment.
+ std::string::const_iterator End = Input.end();
// Multi line comment
if (InComment == true)
{
- for (const char *I = Buffer; *I != 0; I++)
+ for (std::string::const_iterator I = Start;
+ I != End; ++I)
{
- if (*I == '*' && I[1] == '/')
+ if (*I == '*' && I + 1 != End && I[1] == '/')
{
- memmove(Buffer,I+2,strlen(I+2) + 1);
+ Start = I + 2;
InComment = false;
break;
}
@@ -536,57 +574,66 @@
// Discard single line comments
bool InQuote = false;
- for (char *I = Buffer; *I != 0; I++)
+ for (std::string::const_iterator I = Start;
+ I != End; ++I)
{
if (*I == '"')
InQuote = !InQuote;
if (InQuote == true)
continue;
- if (*I == '/' && I[1] == '/')
+ if (*I == '/' && I + 1 != End && I[1] == '/')
{
- *I = 0;
+ End = I;
break;
}
}
- // Look for multi line comments
+ // Look for multi line comments and build up the
+ // fragment.
+ Fragment.reserve(End - Start);
InQuote = false;
- for (char *I = Buffer; *I != 0; I++)
+ for (std::string::const_iterator I = Start;
+ I != End; ++I)
{
if (*I == '"')
InQuote = !InQuote;
if (InQuote == true)
- continue;
-
- if (*I == '/' && I[1] == '*')
+ Fragment.push_back(*I);
+ else if (*I == '/' && I + 1 != End && I[1] == '*')
{
InComment = true;
- for (char *J = Buffer; *J != 0; J++)
+ for (std::string::const_iterator J = I;
+ J != End; ++J)
{
- if (*J == '*' && J[1] == '/')
+ if (*J == '*' && J + 1 != End && J[1] == '/')
{
- memmove(I,J+2,strlen(J+2) + 1);
+ // Pretend we just finished walking over the
+ // comment, and don't add anything to the output
+ // fragment.
+ I = J + 1;
InComment = false;
break;
}
}
if (InComment == true)
- {
- *I = 0;
- break;
- }
+ break;
}
+ else
+ Fragment.push_back(*I);
}
-
- // Blank
- if (Buffer[0] == 0)
+
+ // Skip blank lines.
+ if (Fragment.empty())
continue;
- // We now have a valid line fragment
+ // The line has actual content; interpret what it means.
InQuote = false;
- for (char *I = Buffer; *I != 0;)
+ Start = Fragment.begin();
+ End = Fragment.end();
+ for (std::string::const_iterator I = Start;
+ I != End; ++I)
{
if (*I == '"')
InQuote = !InQuote;
@@ -594,18 +641,21 @@
if (InQuote == false && (*I == '{' || *I == ';' || *I == '}'))
{
// Put the last fragment into the buffer
- char *Start = Buffer;
- char *Stop = I;
- for (; Start != I && isspace(*Start) != 0; Start++);
- for (; Stop != Start && isspace(Stop[-1]) != 0; Stop--);
- if (LineBuffer.empty() == false && Stop - Start != 0)
+ std::string::const_iterator NonWhitespaceStart = Start;
+ std::string::const_iterator NonWhitespaceStop = I;
+ for (; NonWhitespaceStart != I && isspace(*NonWhitespaceStart) != 0; NonWhitespaceStart++)
+ ;
+ for (; NonWhitespaceStop != NonWhitespaceStart && isspace(NonWhitespaceStop[-1]) != 0; NonWhitespaceStop--)
+ ;
+ if (LineBuffer.empty() == false && NonWhitespaceStop - NonWhitespaceStart != 0)
LineBuffer += ' ';
- LineBuffer += string(Start,Stop - Start);
-
- // Remove the fragment
+ LineBuffer += string(NonWhitespaceStart, NonWhitespaceStop);
+
+ // Drop this from the input string, saving the character
+ // that terminated the construct we just closed. (i.e., a
+ // brace or a semicolon)
char TermChar = *I;
- memmove(Buffer,I + 1,strlen(I + 1) + 1);
- I = Buffer;
+ Start = I + 1;
// Syntax Error
if (TermChar == '{' && LineBuffer.empty() == true)
@@ -727,15 +777,32 @@
}
}
- else
- I++;
- }
-
- // Store the fragment
- const char *Stripd = _strstrip(Buffer);
- if (*Stripd != 0 && LineBuffer.empty() == false)
- LineBuffer += " ";
- LineBuffer += Stripd;
+ }
+
+ // Store the remaining text, if any, in the current line buffer.
+
+ // NB: could change this to use string-based operations; I'm
+ // using strstrip now to ensure backwards compatibility.
+ // -- dburrows 2008-04-01
+ {
+ char *Buffer = new char[End - Start + 1];
+ try
+ {
+ std::copy(Start, End, Buffer);
+ Buffer[End - Start] = '\0';
+
+ const char *Stripd = _strstrip(Buffer);
+ if (*Stripd != 0 && LineBuffer.empty() == false)
+ LineBuffer += " ";
+ LineBuffer += Stripd;
+ }
+ catch(...)
+ {
+ delete[] Buffer;
+ throw;
+ }
+ delete[] Buffer;
+ }
}
if (LineBuffer.empty() == false)
