summaryrefslogtreecommitdiff
path: root/apt-pkg/tagfile.cc
diff options
context:
space:
mode:
authorArch Librarian <arch@canonical.com>2004-09-20 16:56:32 +0000
committerArch Librarian <arch@canonical.com>2004-09-20 16:56:32 +0000
commitb2e465d6d32d2dc884f58b94acb7e35f671a87fe (patch)
tree5928383b9bde7b0ba9812e6526ad746466e558f7 /apt-pkg/tagfile.cc
parent00b47c98ca4a4349686a082eba6d77decbb03a4d (diff)
Join with aliencode
Author: jgg Date: 2001-02-20 07:03:16 GMT Join with aliencode
Diffstat (limited to 'apt-pkg/tagfile.cc')
-rw-r--r--apt-pkg/tagfile.cc282
1 files changed, 259 insertions, 23 deletions
diff --git a/apt-pkg/tagfile.cc b/apt-pkg/tagfile.cc
index 97e5c244a..5d1144629 100644
--- a/apt-pkg/tagfile.cc
+++ b/apt-pkg/tagfile.cc
@@ -1,6 +1,6 @@
// -*- mode: cpp; mode: fold -*-
// Description /*{{{*/
-// $Id: tagfile.cc,v 1.25 1999/07/03 06:45:40 jgg Exp $
+// $Id: tagfile.cc,v 1.26 2001/02/20 07:03:17 jgg Exp $
/* ######################################################################
Fast scanner for RFC-822 type header information
@@ -19,6 +19,8 @@
#include <apt-pkg/error.h>
#include <apt-pkg/strutl.h>
+#include <apti18n.h>
+
#include <string>
#include <stdio.h>
/*}}}*/
@@ -26,16 +28,17 @@
// TagFile::pkgTagFile - Constructor /*{{{*/
// ---------------------------------------------------------------------
/* */
-pkgTagFile::pkgTagFile(FileFd &Fd,unsigned long Size) : Fd(Fd), Size(Size)
+pkgTagFile::pkgTagFile(FileFd *pFd,unsigned long Size) : Fd(*pFd), Size(Size)
{
Buffer = new char[Size];
Start = End = Buffer;
Left = Fd.Size();
+ TotalSize = Fd.Size();
iOffset = 0;
Fill();
}
/*}}}*/
-// pkgTagFile::~pkgTagFile - Destructor /*{{{*/
+// TagFile::~pkgTagFile - Destructor /*{{{*/
// ---------------------------------------------------------------------
/* */
pkgTagFile::~pkgTagFile()
@@ -54,10 +57,12 @@ bool pkgTagFile::Step(pkgTagSection &Tag)
return false;
if (Tag.Scan(Start,End - Start) == false)
- return _error->Error("Unable to parse package file %s (1)",Fd.Name().c_str());
+ return _error->Error(_("Unable to parse package file %s (1)"),Fd.Name().c_str());
}
Start += Tag.size();
iOffset += Tag.size();
+
+ Tag.Trim();
return true;
}
@@ -118,8 +123,18 @@ bool pkgTagFile::Fill()
that is there */
bool pkgTagFile::Jump(pkgTagSection &Tag,unsigned long Offset)
{
+ // We are within a buffer space of the next hit..
+ if (Offset >= iOffset && iOffset + (End - Start) > Offset)
+ {
+ unsigned long Dist = Offset - iOffset;
+ Start += Dist;
+ iOffset += Dist;
+ return Step(Tag);
+ }
+
+ // Reposition and reload..
iOffset = Offset;
- Left = Fd.Size() - Offset;
+ Left = TotalSize - Offset;
if (Fd.Seek(Offset) == false)
return false;
End = Start = Buffer;
@@ -135,10 +150,7 @@ bool pkgTagFile::Jump(pkgTagSection &Tag,unsigned long Offset)
return false;
if (Tag.Scan(Start,End - Start) == false)
- {
- cout << string(Start,End) << endl;
- return _error->Error("Unable to parse package file %s (2)",Fd.Name().c_str());
- }
+ return _error->Error(_("Unable to parse package file %s (2)"),Fd.Name().c_str());
return true;
}
@@ -148,6 +160,14 @@ bool pkgTagFile::Jump(pkgTagSection &Tag,unsigned long Offset)
/* This looks for the first double new line in the data stream. It also
indexes the tags in the section. This very simple hash function for the
first 3 letters gives very good performance on the debian package files */
+inline static unsigned long AlphaHash(const char *Text, const char *End = 0)
+{
+ unsigned long Res = 0;
+ for (; Text != End && *Text != ':' && *Text != 0; Text++)
+ Res = (unsigned long)(*Text) ^ (Res << 2);
+ return Res & 0xFF;
+}
+
bool pkgTagSection::Scan(const char *Start,unsigned long MaxLength)
{
const char *End = Start + MaxLength;
@@ -164,10 +184,7 @@ bool pkgTagSection::Scan(const char *Start,unsigned long MaxLength)
if (isspace(Stop[0]) == 0)
{
Indexes[TagCount++] = Stop - Section;
- unsigned char A = tolower(Stop[0]) - 'a';
- unsigned char B = tolower(Stop[1]) - 'a';
- unsigned char C = tolower(Stop[3]) - 'a';
- AlphaIndexes[((A + C/3)%26) + 26*((B + C/2)%26)] = TagCount;
+ AlphaIndexes[AlphaHash(Stop,End)] = TagCount;
}
Stop = (const char *)memchr(Stop,'\n',End - Stop);
@@ -191,17 +208,21 @@ bool pkgTagSection::Scan(const char *Start,unsigned long MaxLength)
return false;
}
/*}}}*/
+// TagSection::Trim - Trim off any trailing garbage /*{{{*/
+// ---------------------------------------------------------------------
+/* There should be exactly 1 newline at the end of the buffer, no more. */
+void pkgTagSection::Trim()
+{
+ for (; Stop > Section + 2 && (Stop[-2] == '\n' || Stop[-2] == '\r'); Stop--);
+}
+ /*}}}*/
// TagSection::Find - Locate a tag /*{{{*/
// ---------------------------------------------------------------------
/* This searches the section for a tag that matches the given string. */
-bool pkgTagSection::Find(const char *Tag,const char *&Start,
- const char *&End)
+bool pkgTagSection::Find(const char *Tag,unsigned &Pos) const
{
unsigned int Length = strlen(Tag);
- unsigned char A = tolower(Tag[0]) - 'a';
- unsigned char B = tolower(Tag[1]) - 'a';
- unsigned char C = tolower(Tag[3]) - 'a';
- unsigned int I = AlphaIndexes[((A + C/3)%26) + 26*((B + C/2)%26)];
+ unsigned int I = AlphaIndexes[AlphaHash(Tag)];
if (I == 0)
return false;
I--;
@@ -219,6 +240,39 @@ bool pkgTagSection::Find(const char *Tag,const char *&Start,
for (; isspace(*C) != 0; C++);
if (*C != ':')
continue;
+ Pos = I;
+ return true;
+ }
+
+ Pos = 0;
+ return false;
+}
+ /*}}}*/
+// TagSection::Find - Locate a tag /*{{{*/
+// ---------------------------------------------------------------------
+/* This searches the section for a tag that matches the given string. */
+bool pkgTagSection::Find(const char *Tag,const char *&Start,
+ const char *&End) const
+{
+ unsigned int Length = strlen(Tag);
+ unsigned int I = AlphaIndexes[AlphaHash(Tag)];
+ if (I == 0)
+ return false;
+ I--;
+
+ for (unsigned int Counter = 0; Counter != TagCount; Counter++,
+ I = (I+1)%TagCount)
+ {
+ const char *St;
+ St = Section + Indexes[I];
+ if (strncasecmp(Tag,St,Length) != 0)
+ continue;
+
+ // Make sure the colon is in the right place
+ const char *C = St + Length;
+ for (; isspace(*C) != 0; C++);
+ if (*C != ':')
+ continue;
// Strip off the gunk from the start end
Start = C;
@@ -239,7 +293,7 @@ bool pkgTagSection::Find(const char *Tag,const char *&Start,
// TagSection::FindS - Find a string /*{{{*/
// ---------------------------------------------------------------------
/* */
-string pkgTagSection::FindS(const char *Tag)
+string pkgTagSection::FindS(const char *Tag) const
{
const char *Start;
const char *End;
@@ -251,7 +305,7 @@ string pkgTagSection::FindS(const char *Tag)
// TagSection::FindI - Find an integer /*{{{*/
// ---------------------------------------------------------------------
/* */
-signed int pkgTagSection::FindI(const char *Tag,signed long Default)
+signed int pkgTagSection::FindI(const char *Tag,signed long Default) const
{
const char *Start;
const char *Stop;
@@ -276,7 +330,7 @@ signed int pkgTagSection::FindI(const char *Tag,signed long Default)
// ---------------------------------------------------------------------
/* The bits marked in Flag are masked on/off in Flags */
bool pkgTagSection::FindFlag(const char *Tag,unsigned long &Flags,
- unsigned long Flag)
+ unsigned long Flag) const
{
const char *Start;
const char *Stop;
@@ -294,9 +348,191 @@ bool pkgTagSection::FindFlag(const char *Tag,unsigned long &Flags,
return true;
default:
- _error->Warning("Unknown flag value");
+ _error->Warning("Unknown flag value: %s",string(Start,Stop).c_str());
return true;
}
return true;
}
/*}}}*/
+
+// TFRewrite - Rewrite a control record /*{{{*/
+// ---------------------------------------------------------------------
+/* This writes the control record to stdout rewriting it as necessary. The
+ override map item specificies the rewriting rules to follow. This also
+ takes the time to sort the feild list. */
+
+/* The order of this list is taken from dpkg source lib/parse.c the fieldinfos
+ array. */
+static const char *iTFRewritePackageOrder[] = {
+ "Package",
+ "Essential",
+ "Status",
+ "Priority",
+ "Section",
+ "Installed-Size",
+ "Maintainer",
+ "Architecture",
+ "Source",
+ "Version",
+ "Revision", // Obsolete
+ "Config-Version", // Obsolete
+ "Replaces",
+ "Provides",
+ "Depends",
+ "Pre-Depends",
+ "Recommends",
+ "Suggests",
+ "Conflicts",
+ "Conffiles",
+ "Filename",
+ "Size",
+ "MD5Sum",
+ "MSDOS-Filename", // Obsolete
+ "Description",
+ 0};
+static const char *iTFRewriteSourceOrder[] = {"Package",
+ "Source",
+ "Binary",
+ "Version",
+ "Priority",
+ "Section",
+ "Maintainer",
+ "Build-Depends",
+ "Build-Depends-Indep",
+ "Build-Conflicts",
+ "Build-Conflicts-Indep",
+ "Architecture",
+ "Standards-Version",
+ "Format",
+ "Directory",
+ "Files",
+ 0};
+
+/* Two levels of initialization are used because gcc will set the symbol
+ size of an array to the length of the array, causing dynamic relinking
+ errors. Doing this makes the symbol size constant */
+const char **TFRewritePackageOrder = iTFRewritePackageOrder;
+const char **TFRewriteSourceOrder = iTFRewriteSourceOrder;
+
+bool TFRewrite(FILE *Output,pkgTagSection const &Tags,const char *Order[],
+ TFRewriteData *Rewrite)
+{
+ unsigned char Visited[256]; // Bit 1 is Order, Bit 2 is Rewrite
+ for (unsigned I = 0; I != 256; I++)
+ Visited[I] = 0;
+
+ // Set new tag up as necessary.
+ for (unsigned int J = 0; Rewrite != 0 && Rewrite[J].Tag != 0; J++)
+ {
+ if (Rewrite[J].NewTag == 0)
+ Rewrite[J].NewTag = Rewrite[J].Tag;
+ }
+
+ // Write all all of the tags, in order.
+ for (unsigned int I = 0; Order[I] != 0; I++)
+ {
+ bool Rewritten = false;
+
+ // See if this is a field that needs to be rewritten
+ for (unsigned int J = 0; Rewrite != 0 && Rewrite[J].Tag != 0; J++)
+ {
+ if (strcasecmp(Rewrite[J].Tag,Order[I]) == 0)
+ {
+ Visited[J] |= 2;
+ if (Rewrite[J].Rewrite != 0 && Rewrite[J].Rewrite[0] != 0)
+ {
+ if (isspace(Rewrite[J].Rewrite[0]))
+ fprintf(Output,"%s:%s\n",Rewrite[J].NewTag,Rewrite[J].Rewrite);
+ else
+ fprintf(Output,"%s: %s\n",Rewrite[J].NewTag,Rewrite[J].Rewrite);
+ }
+
+ Rewritten = true;
+ break;
+ }
+ }
+
+ // See if it is in the fragment
+ unsigned Pos;
+ if (Tags.Find(Order[I],Pos) == false)
+ continue;
+ Visited[Pos] |= 1;
+
+ if (Rewritten == true)
+ continue;
+
+ /* Write out this element, taking a moment to rewrite the tag
+ in case of changes of case. */
+ const char *Start;
+ const char *Stop;
+ Tags.Get(Start,Stop,Pos);
+
+ if (fputs(Order[I],Output) < 0)
+ return _error->Errno("fputs","IO Error to output");
+ Start += strlen(Order[I]);
+ if (fwrite(Start,Stop - Start,1,Output) != 1)
+ return _error->Errno("fwrite","IO Error to output");
+ if (Stop[-1] != '\n')
+ fprintf(Output,"\n");
+ }
+
+ // Now write all the old tags that were missed.
+ for (unsigned int I = 0; I != Tags.Count(); I++)
+ {
+ if ((Visited[I] & 1) == 1)
+ continue;
+
+ const char *Start;
+ const char *Stop;
+ Tags.Get(Start,Stop,I);
+ const char *End = Start;
+ for (; End < Stop && *End != ':'; End++);
+
+ // See if this is a field that needs to be rewritten
+ bool Rewritten = false;
+ for (unsigned int J = 0; Rewrite != 0 && Rewrite[J].Tag != 0; J++)
+ {
+ if (stringcasecmp(Start,End,Rewrite[J].Tag) == 0)
+ {
+ Visited[J] |= 2;
+ if (Rewrite[J].Rewrite != 0 && Rewrite[J].Rewrite[0] != 0)
+ {
+ if (isspace(Rewrite[J].Rewrite[0]))
+ fprintf(Output,"%s:%s\n",Rewrite[J].NewTag,Rewrite[J].Rewrite);
+ else
+ fprintf(Output,"%s: %s\n",Rewrite[J].NewTag,Rewrite[J].Rewrite);
+ }
+
+ Rewritten = true;
+ break;
+ }
+ }
+
+ if (Rewritten == true)
+ continue;
+
+ // Write out this element
+ if (fwrite(Start,Stop - Start,1,Output) != 1)
+ return _error->Errno("fwrite","IO Error to output");
+ if (Stop[-1] != '\n')
+ fprintf(Output,"\n");
+ }
+
+ // Now write all the rewrites that were missed
+ for (unsigned int J = 0; Rewrite != 0 && Rewrite[J].Tag != 0; J++)
+ {
+ if ((Visited[J] & 2) == 2)
+ continue;
+
+ if (Rewrite[J].Rewrite != 0 && Rewrite[J].Rewrite[0] != 0)
+ {
+ if (isspace(Rewrite[J].Rewrite[0]))
+ fprintf(Output,"%s:%s\n",Rewrite[J].NewTag,Rewrite[J].Rewrite);
+ else
+ fprintf(Output,"%s: %s\n",Rewrite[J].NewTag,Rewrite[J].Rewrite);
+ }
+ }
+
+ return true;
+}
+ /*}}}*/