1. preprocessor doesn't touch unchanged files anymore

2. doublechedk on whether file is up to date when making
3. changed the way preprocessor targets depend on the timestamp file
4. a bugfix in strip_suffix/strip_prefix

5 files changed, 95 insertions, 22 deletions

diff --git a/include/sitecing/sitecing_enflesher.h b/include/sitecing/sitecing_enflesher.h
index 512a358..8bc43a0 100644
--- a/include/sitecing/sitecing_enflesher.h
+++ b/include/sitecing/sitecing_enflesher.h
@@ -29,38 +29,51 @@ class sitecing_enflesher : public sitecing_enflesherFlexLexer {
 	 * It is time to anchor output with the #line directive.
 	 */
 	bool anchor_time;
 	/**
 	 * The file currently being written is supposed to have #line
 	 * directives all around the place.
 	 */
 	bool anchoraged;
 	/**
 	 * The reference to the parser object containg the parsed source.
 	 */
 	sitecing_parser& parser;
 	/**
 	 * The output stream.
 	 */
 	ofstream outs;
+	/**
+	 * the outs stream destination file.
+	 * @see outs
+	 */
+	string outs_filename;
 
 	/**
 	 * @param p The parser object containing preparsed data.
 	 */
 	sitecing_enflesher(sitecing_parser& p)
 	    : parser(p), anchor_time(true) { }
 
 	/**
 	 * Do the job.
 	 */
 	void enflesh();
 
 	virtual void LexerOutput(const char *buf,int size);
 	virtual int yylex();
 
 	/**
 	 * Put a #line anchor into output.
 	 */
 	void anchor();
+
+	/**
+	 * Close previously opened output stream, rename to the 'correct'
+	 * destination filename, if needed, and open new file.
+	 * @see outs
+	 * @see outs_filename
+	 */
+	void outs_open(const string& nfile);
 };
 
 #endif /* __SITECING_SITECING_ENFLESHER_H */
diff --git a/lib/component_factory.cc b/lib/component_factory.cc
index f8666dc..b8f5a16 100644
--- a/lib/component_factory.cc
+++ b/lib/component_factory.cc
@@ -6,33 +6,33 @@
  #include <unistd.h>
  #include <sys/wait.h>
  #include <fcntl.h>
  #include <iostream>
  #include <fstream>
  #include <stdexcept>
  #include <vector>
  using namespace std;
  #include "sitecing/component_factory.h"
  #include "sitecing/sitecing_util.h"
  #include "sitecing/sitecing_parser.h"
  #include "sitecing/sitecing_exception.h"
 #endif
  
 namespace sitecing {
 
-    static const char *pp_targets[] = { ".cc", ".h", ".imports", ".classname", ".baseclassname", ".ancestors" };
+    static const char *pp_targets[] = { ".cc", ".h", ".imports", ".classname", ".baseclassname", ".ancestors", ".pp_stamp" };
     static const char *cc_targets[] = { ".o", ".d" };
 
     component_factory::component_factory(configuration& c)
 	: config(c),
     root_source(normalize_path(c.root_source,strip_trailing_slash)+'/'),
     root_intermediate(normalize_path(c.root_intermediate,strip_trailing_slash)+'/'),
     root_so(normalize_path(c.root_so,strip_trailing_slash)+'/') {
     }
 
     void component_factory::get_dependencies(const string& dst,file_list_t& deps) {
 	deps.clear();
 	string dp = normalize_path(dst,strip_trailing_slash);
 	// source documents
 	try { // XXX: or just compare it off?
 	    string noro = strip_prefix(dp,root_source);
 	    return;
@@ -103,41 +103,46 @@ namespace sitecing {
 			    if(str.find_first_of("\\:")==string::npos)
 				deps.push_back(combine_path(config.root_source+noro,str));
 			}
 		    }
 		}
 		// XXX: extra deps like IntermediateDeps?
 	    }catch(utility_no_affix& una) {
 		// do nothing.
 	    }
 	}
     }
 
     bool component_factory::is_uptodate(const string& dst,file_list_t *deps) {
 	string dp = normalize_path(dst,strip_trailing_slash);
 	// XXX: or just compare it off, instead of throwing things around.
 	try {
-	    strip_prefix(dp,root_intermediate);
-	    return file_factory::is_uptodate(dst,deps);
-	}catch(utility_no_prefix& unp) {
-	}
+	    string noro = strip_prefix(dp,root_intermediate);
+	    for(int ppt=0;(ppt+1)<sizeof(pp_targets)/sizeof(*pp_targets);ppt++) {
+		try {
+		    string nos = strip_suffix(noro,pp_targets[ppt]);
+		    return file_factory::is_uptodate(root_intermediate+nos+".pp_stamp",deps);
+		}catch(utility_no_suffix& uns) { }
+	    }
+	    bool rv =  file_factory::is_uptodate(dst,deps);
+	    return rv;
+	}catch(utility_no_prefix& unp) { }
 	try {
 	    strip_prefix(dp,root_so);
 	    return file_factory::is_uptodate(dst,deps);
-	}catch(utility_no_prefix& unp) {
-	}
+	}catch(utility_no_prefix& unp) { }
 	return true;
     }
 
     void component_factory::build(const string& dst) {
 	string dp = normalize_path(dst,strip_trailing_slash);
 	// sources
 	try {
 	    string noro = strip_prefix(dp,root_source);
 	    // building the sources is left up to developer
 	    return;
 	}catch(utility_no_prefix& unp) {
 	}
 	// .so files
 	try {
 	    string noso = strip_suffix(dp,".so");
 	    string noro = strip_prefix(noso,root_so);
@@ -236,32 +241,34 @@ namespace sitecing {
 		sitecing_parser parser(*this);
 		config_options *co_skeleton = config.lookup_config(noro,config_options::flag_skeleton);
 		if(co_skeleton)
 		    parser.skeleton = co_skeleton->skeleton;
 		static const char *id_chars = "abcdefghijklmnopqrstuvwxyz0123456789_ABCDEFGHIJKLMNOPQRSTUVWXYZ";
 		parser.class_name = normalize_path(noro,strip_leading_slash|strip_trailing_slash);
 		for(string::size_type illc = parser.class_name.find_first_not_of(id_chars);illc!=string::npos;illc=parser.class_name.find_first_not_of(id_chars,illc)) {
 		    string::size_type lc = parser.class_name.find_first_of(id_chars,illc);
 		    int n = ((lc==string::npos)?parser.class_name.length():lc)-illc;
 		    parser.class_name.replace(illc,n,n,'_');
 		}
 		parser.class_name = "_SCC_"+parser.class_name;
 		parser.output_basename = nos;
 		parser.component_basename = noro;
 		try {
 		    parser.preprocess(src);
+		    string sf = root_intermediate+noro+".pp_stamp";
+		    ofstream sfs(sf.c_str(),ios::trunc|ios::out); // touch .pp_stamp
 		}catch(preprocessor_error& pe) {
 		    pe.component_name = noro;
 		    pe.see(CODEPOINT);
 		    throw;
 		}
 		return;
 	    }catch(utility_no_affix& una) {
 		// must be a crap from .d file
 	    }
 	}
 	cerr << "ignoring build request for " << dp << endl;
     }
 
     int component_factory::execute(const string& cmd, const list<string>& args,int stdo,int stde) {
 	// XXX: is it right that we do stdio/stderr tricks outside of the function?
 	cerr << "executing: " << cmd;
diff --git a/lib/file_factory.cc b/lib/file_factory.cc
index c6b5748..7ca7b86 100644
--- a/lib/file_factory.cc
+++ b/lib/file_factory.cc
@@ -26,30 +26,31 @@ namespace sitecing {
 		return false;
 	    if(!is_uptodate(*i))
 		return false;
 	}
 	return true;
     }
 
     void file_factory::make(const string& dst) {
 	try {
 	    depth++;
 	    if(depth>25)
 		throw konforka::exception(CODEPOINT,"recursed too deeply.");
 	    file_list_t deps;
 	    if(!is_uptodate(dst,&deps)) {
 		for(file_list_t::const_iterator i=deps.begin();i!=deps.end();i++)
 		    make(*i);
-		build(dst);
+		if(!is_uptodate(dst,&deps))
+		    build(dst);
 	    }
 	    depth--;
 	}catch(konforka::exception& ke) {
 	    depth--;
 	    ke.see(CODEPOINT);
 	    throw;
 	}catch(...) {
 	    depth--;
 	    throw;
 	}
     }
 
 }
diff --git a/lib/sitecing_enflesher.ll b/lib/sitecing_enflesher.ll
index 5f631d7..46489c7 100644
--- a/lib/sitecing_enflesher.ll
+++ b/lib/sitecing_enflesher.ll
@@ -9,51 +9,41 @@ using namespace sitecing;
 #define sitecing_enflesher_flexlexer_once
 #include "sitecing/sitecing_enflesher.h"
 #include "sitecing/sitecing_parser.h"
 #undef yyFlexLexer
 #define yyFlexLexer sitecing_enflesherFlexLexer
 %}
 %option 8bit c++ verbose noyywrap yyclass="sitecing_enflesher" yylineno prefix="sitecing_enflesher" stack debug
 
 ID	[A-Za-z_][A-Za-z0-9_]*
 
 %%
 
 ^\%\%\#[^\n]+\n	{
     string line = yytext;
     line.erase(0,3);
     line.erase(line.length()-1);
-    outs.flush();
-    outs.close();
-    outs.clear();
-    outs.open((parser.output_basename+line).c_str(),ios::trunc);
-    if(!outs.good())
-	throw preprocessor_error(CODEPOINT,"failed to write preprocessor output");
+    outs_open(parser.output_basename+line);
     anchor();
     anchoraged = true;
 }
 ^\%\%[^\n]+\n	{
     string line = yytext;
     line.erase(0,2);
     line.erase(line.length()-1);
-    outs.flush();
-    outs.close();
-    outs.clear();
-    outs.open((parser.output_basename+line).c_str(),ios::trunc);
-    if(!outs.good())
-	throw preprocessor_error(CODEPOINT,"failed to write preprocessor output");
+    outs_open(parser.output_basename+line);
     anchoraged = false;
 }
 
 \<\%component_basename\%\>	outs << parser.component_basename; anchor_time = true;
 \<\%impl\%\>			outs << parser.impl; anchor_time = true;
 \<\%member_functions:impl\%\>	{
     for(sitecing_parser::member_functions_t::const_iterator i=parser.member_functions.begin();i!=parser.member_functions.end();i++) {
 	outs << i->type << " " << parser.class_name << "::";
 	if(i->name.empty()) {
 	    outs << parser.class_name << "()";
 	    bool first = true;
 	    for(sitecing_parser::member_variables_t::iterator i=parser.member_variables.begin();i!=parser.member_variables.end();i++) {
 		if(i->initializer.empty())
 		    continue;
 		if(first) {
 		    outs << ":";
@@ -184,19 +174,81 @@ void sitecing_enflesher::LexerOutput(const char *buf,int size) {
 }
 
 void sitecing_enflesher::enflesh() {
     ifstream ifs(parser.skeleton.c_str());
     if(!ifs.good())
 	throw preprocessor_error(CODEPOINT,"failed to open skeleton file");
     switch_streams(&ifs,NULL);
     yylex();
 }
 
 void sitecing_enflesher::anchor() {
     if(!anchoraged)
 	return;
     outs << "\n#line " << lineno() << " \"" << parser.skeleton << "\"\n";
     anchor_time = false;
 }
+
+void sitecing_enflesher::outs_open(const string& nfile) {
+    if(!outs_filename.empty()) {
+	outs.flush();
+	outs.close();
+	outs.clear();
+	/*
+	 * compare source and destination files.
+	 *
+	 * one can also keep a hash for the old one and compute one for the
+	 * output while writing, but I'm not sure if it's any better. Surely a
+	 * bit less accurate, unless we're going to compare it in case of
+	 * difference, anyway.
+	 */
+	bool overwrite = false;
+	struct stat st_s, st_d;
+	string fn_s = outs_filename+".new";
+	string fn_d = outs_filename;
+	if(stat(fn_d.c_str(),&st_d)) {
+	    overwrite = true;
+	}else{
+	    if(stat(fn_s.c_str(),&st_s))
+		throw preprocessor_error(CODEPOINT,"failed to stat() supposedly created file");
+	    if(st_s.st_size!=st_d.st_size) {
+		overwrite = true;
+	    }else{
+		ifstream i_s(fn_s.c_str(),ios::in);
+		if(!i_s)
+		    throw preprocessor_error(CODEPOINT,"failed to open supposedly created file");
+		ifstream i_d(fn_d.c_str(),ios::in);
+		if(!i_d)
+		    throw preprocessor_error(CODEPOINT,"failed to open the old preprocessed source");
+		off_t remaining = st_s.st_size;
+		char t1[2048];
+		char t2[sizeof(t1)];
+		while(remaining) {
+		    int rb = remaining;
+		    if(rb>sizeof(t1))
+			rb = sizeof(t1);
+		    if(i_s.read(t1,rb).gcount()!=rb)
+			throw preprocessor_error(CODEPOINT,"error reading just created file");
+		    if(i_d.read(t2,rb).gcount()!=rb)
+			throw preprocessor_error(CODEPOINT,"error reading the old preprocessed source");
+		    if(memcmp(t1,t2,rb)) {
+			overwrite = true;
+			break;
+		    }
+		    remaining -= rb;
+		}
+	    }
+	}
+	if(overwrite) {
+	    cerr << "renaming '" << fn_s << "'" << endl;
+	    if(rename(fn_s.c_str(),fn_d.c_str()))
+		throw preprocessor_error(CODEPOINT,"failed to rename() generated output");
+	}
+    }
+    outs_filename = nfile;
+    outs.open((nfile+".new").c_str(),ios::trunc);
+    if(!outs.good())
+	throw preprocessor_error(CODEPOINT,"failed to write preprocessor output");
+}
 /*
  * vim:set ft=lex:
  */
diff --git a/lib/sitecing_util.cc b/lib/sitecing_util.cc
index 5466b28..f892a60 100644
--- a/lib/sitecing_util.cc
+++ b/lib/sitecing_util.cc
@@ -43,39 +43,39 @@ namespace sitecing {
 			|| ( s[0]=='/' && s[1]=='.' && s[2]=='.' && (s[3]==0 || s[3]=='/') ) // "/..(/|$)"
 		  )
 		    throw utility_restricted_sequence(CODEPOINT,"restricted updir sequence encountered");
 	    }
 	    rv += *s;
 	    if( (*s) != '/' )
 		notslash=rv.length();
 	}
 	if(!(opts&strip_trailing_slash))
 	    notslash++;
 	if(notslash<rv.length())
 	    rv.erase(notslash); // XXX: check the logic of stripping/not strippling trailing slash
 	return rv;
     }
 
     string strip_prefix(const string& str,const string& prefix) {
-	if(str.compare(0,prefix.length(),prefix))
+	if( (str.length()<prefix.length()) || str.compare(0,prefix.length(),prefix))
 	    throw utility_no_prefix(CODEPOINT,"no such prefix");
 	return str.substr(prefix.length());
     }
 
     string strip_suffix(const string& str,const string& suffix) {
-	if(str.compare(str.length()-suffix.length(),suffix.length(),suffix))
+	if( (str.length()<suffix.length()) || str.compare(str.length()-suffix.length(),suffix.length(),suffix))
 	    throw utility_no_suffix(CODEPOINT,"no such suffix");
 	return str.substr(0,str.length()-suffix.length());
     }
 
     string dir_name(const string& filename) {
 	string::size_type sl = filename.find_last_of('/');
 	if(sl==string::npos)
 	    return ""; // no slashes -- no dir.
 	string::size_type nosl = filename.find_last_not_of('/',sl);
 	if(nosl==string::npos)
 	    return ""; // only slashes -- no dir. XXX: only slashes after the last slash... does it mean no dir?
 	return filename.substr(0,nosl+1);
     }
 
     void make_path(const string& path,mode_t mode) {
 	struct stat st;