Extract a Tar - Tar Extractor
Some performance measurements showed that zlib performed quite poorly
- in comparision to a forked gzip process. This tar extractor makes use
+ in comparison to a forked gzip process. This tar extractor makes use
of the fact that dup'd file descriptors have the same seek pointer
and that gzip will not read past the end of a compressed stream,
even if there is more data. We use the dup property to track extraction
#include <apt-pkg/error.h>
#include <apt-pkg/strutl.h>
#include <apt-pkg/configuration.h>
-#include <apt-pkg/macros.h>
+#include <apt-pkg/fileutl.h>
-#include <stdlib.h>
+#include <string.h>
+#include <algorithm>
+#include <string>
#include <unistd.h>
#include <signal.h>
#include <fcntl.h>
// ExtractTar::ExtractTar - Constructor /*{{{*/
// ---------------------------------------------------------------------
/* */
-ExtractTar::ExtractTar(FileFd &Fd,unsigned long Max,string DecompressionProgram) : File(Fd),
- MaxInSize(Max), DecompressProg(DecompressionProgram)
-
+ExtractTar::ExtractTar(FileFd &Fd,unsigned long long Max,string DecompressionProgram)
+ : File(Fd), MaxInSize(Max), DecompressProg(DecompressionProgram)
{
GZPid = -1;
- InFd = -1;
Eof = false;
}
/*}}}*/
gzip will efficiently ignore the extra bits. */
bool ExtractTar::StartGzip()
{
+ if (DecompressProg.empty())
+ {
+ InFd.OpenDescriptor(File.Fd(), FileFd::ReadOnly, FileFd::None, false);
+ return true;
+ }
+
int Pipes[2];
if (pipe(Pipes) != 0)
return _error->Errno("pipe",_("Failed to create pipes"));
-
+
// Fork off the process
GZPid = ExecFork();
dup2(Fd,STDERR_FILENO);
close(Fd);
SetCloseExec(STDOUT_FILENO,false);
- SetCloseExec(STDIN_FILENO,false);
+ SetCloseExec(STDIN_FILENO,false);
SetCloseExec(STDERR_FILENO,false);
-
+
const char *Args[3];
string confvar = string("dir::bin::") + DecompressProg;
string argv0 = _config->Find(confvar.c_str(),DecompressProg.c_str());
}
// Fix up our FDs
- InFd.Fd(Pipes[0]);
+ InFd.OpenDescriptor(Pipes[0], FileFd::ReadOnly, FileFd::None, true);
close(Pipes[1]);
return true;
}
return false;
// Loop over all blocks
- string LastLongLink;
- string LastLongName;
+ string LastLongLink, ItemLink;
+ string LastLongName, ItemName;
while (1)
{
bool BadRecord = false;
StrToNum(Tar->Major,Itm.Major,sizeof(Tar->Major),8) == false ||
StrToNum(Tar->Minor,Itm.Minor,sizeof(Tar->Minor),8) == false)
return _error->Error(_("Corrupted archive"));
-
- // Grab the filename
+
+ // Grab the filename and link target: use last long name if one was
+ // set, otherwise use the header value as-is, but remember that it may
+ // fill the entire 100-byte block and needs to be zero-terminated.
+ // See Debian Bug #689582.
if (LastLongName.empty() == false)
Itm.Name = (char *)LastLongName.c_str();
else
- {
- Tar->Name[sizeof(Tar->Name)-1] = 0;
- Itm.Name = Tar->Name;
- }
+ Itm.Name = (char *)ItemName.assign(Tar->Name, sizeof(Tar->Name)).c_str();
if (Itm.Name[0] == '.' && Itm.Name[1] == '/' && Itm.Name[2] != 0)
Itm.Name += 2;
-
- // Grab the link target
- Tar->Name[sizeof(Tar->LinkName)-1] = 0;
- Itm.LinkTarget = Tar->LinkName;
if (LastLongLink.empty() == false)
Itm.LinkTarget = (char *)LastLongLink.c_str();
-
+ else
+ Itm.LinkTarget = (char *)ItemLink.assign(Tar->LinkName, sizeof(Tar->LinkName)).c_str();
+
// Convert the type over
switch (Tar->LinkFlag)
{
case GNU_LongLink:
{
- unsigned long Length = Itm.Size;
+ unsigned long long Length = Itm.Size;
unsigned char Block[512];
while (Length > 0)
{
case GNU_LongName:
{
- unsigned long Length = Itm.Size;
+ unsigned long long Length = Itm.Size;
unsigned char Block[512];
while (Length > 0)
{
return false;
// Copy the file over the FD
- unsigned long Size = Itm.Size;
+ unsigned long long Size = Itm.Size;
while (Size != 0)
{
unsigned char Junk[32*1024];
- unsigned long Read = min(Size,(unsigned long)sizeof(Junk));
+ unsigned long Read = min(Size, (unsigned long long)sizeof(Junk));
if (InFd.Read(Junk,((Read+511)/512)*512) == false)
return false;