mirror of
https://github.com/fadden/nulib2.git
synced 2024-12-29 15:29:16 +00:00
132a8338b9
Some fixes to the Mac OS X build: - Replace the Carbon calls that were used to set the creator and file type with xattr calls. The Carbon stuff still worked but caused deprecation warnings. Stop linking against Carbon. - Correct the way resource forks are accessed (from "/rsrc" to "/..namedfork/rsrc"). The native resource fork support is incomplete and doesn't work quite right, so it's now disabled. (Which means the corrections to the file name don't actually do anything, but you can at least play with it.) - Correct the file/aux type conversion, which appeared to do useful things but actually didn't in some circumstances (e.g. when adding files, the code for acquiring the file types needs to be in NuLib2, not NufxLib). - Set creator and file type to 'pdos' values when extracting from a Binary ][ archive. Also, drop some old purify/quantify stuff.
1449 lines
42 KiB
C
1449 lines
42 KiB
C
/*
|
|
* NuFX archive manipulation library
|
|
* Copyright (C) 2000-2007 by Andy McFadden, All Rights Reserved.
|
|
* This is free software; you can redistribute it and/or modify it under the
|
|
* terms of the BSD License, see the file COPYING-LIB.
|
|
*
|
|
* Operations on output (i.e. non-archive) files, largely system-specific.
|
|
* Portions taken from NuLib, including some code that Devin Reade worked on.
|
|
*
|
|
* It could be argued that "create file" should be a callback function,
|
|
* since it is so heavily system-specific, and most of the other
|
|
* system dependencies are handled by the application rather than the
|
|
* NuFX library. It would also provide a cleaner solution for renaming
|
|
* extracted files. However, the goal of the library is to do the work
|
|
* for the application, not the other way around; and while it might be
|
|
* nice to offload all direct file handling on the application, it
|
|
* complicates rather than simplifies the interface.
|
|
*/
|
|
#include "NufxLibPriv.h"
|
|
|
|
#ifdef MAC_LIKE
|
|
# include <sys/xattr.h>
|
|
#endif
|
|
|
|
/*
|
|
* For systems (e.g. Visual C++ 6.0) that don't have these standard values.
|
|
*/
|
|
#ifndef S_IRUSR
|
|
# define S_IRUSR 0400
|
|
# define S_IWUSR 0200
|
|
# define S_IXUSR 0100
|
|
# define S_IRWXU (S_IRUSR|S_IWUSR|S_IXUSR)
|
|
# define S_IRGRP (S_IRUSR >> 3)
|
|
# define S_IWGRP (S_IWUSR >> 3)
|
|
# define S_IXGRP (S_IXUSR >> 3)
|
|
# define S_IRWXG (S_IRWXU >> 3)
|
|
# define S_IROTH (S_IRGRP >> 3)
|
|
# define S_IWOTH (S_IWGRP >> 3)
|
|
# define S_IXOTH (S_IXGRP >> 3)
|
|
# define S_IRWXO (S_IRWXG >> 3)
|
|
#endif
|
|
#ifndef S_ISREG
|
|
# define S_ISREG(m) (((m) & S_IFMT) == S_IFREG)
|
|
# define S_ISDIR(m) (((m) & S_IFMT) == S_IFDIR)
|
|
#endif
|
|
|
|
/*
|
|
* ===========================================================================
|
|
* DateTime conversions
|
|
* ===========================================================================
|
|
*/
|
|
|
|
/*
|
|
* Dates and times in a NuFX archive are always considered to be in the
|
|
* local time zone. The use of GMT time stamps would have been more
|
|
* appropriate for an archive, but local time works well enough.
|
|
*
|
|
* Regarding Y2K on the Apple II:
|
|
*
|
|
* Dave says P8 drivers should return year values in the range 0..99, where
|
|
* 40..99 = 1940..1999, and 0..39 = 2000..2039. Year values 100..127 should
|
|
* never be used. For ProDOS 8, the year 2000 is "00".
|
|
*
|
|
* The IIgs ReadTimeHex call uses "year minus 1900". For GS/OS, the year
|
|
* 2000 is "100".
|
|
*
|
|
* The NuFX file type note says the archive format should work like
|
|
* The IIgs ReadTimeHex call, which uses "year minus 1900" as its
|
|
* format. GS/ShrinkIt v1.1 uses the IIgs date calls, and so stores the
|
|
* year 2000 as "100". P8 ShrinkIt v3.4 uses the P8 mapping, and stores
|
|
* it as "0". Neither really quite understands what the other is doing.
|
|
*
|
|
* For our purposes, we will follow the NuFX standard and emit "100"
|
|
* for the year 2000, but will accept and understand "0" as well.
|
|
*/
|
|
|
|
|
|
#if defined(UNIX_LIKE) || defined(WINDOWS_LIKE)
|
|
/*
|
|
* Convert from local time in a NuDateTime struct to GMT seconds since 1970.
|
|
*
|
|
* If the conversion is invalid, "*pWhen" is set to zero.
|
|
*/
|
|
static void Nu_DateTimeToGMTSeconds(const NuDateTime* pDateTime, time_t* pWhen)
|
|
{
|
|
struct tm tmbuf;
|
|
time_t when;
|
|
|
|
Assert(pDateTime != NULL);
|
|
Assert(pWhen != NULL);
|
|
|
|
tmbuf.tm_sec = pDateTime->second;
|
|
tmbuf.tm_min = pDateTime->minute;
|
|
tmbuf.tm_hour = pDateTime->hour;
|
|
tmbuf.tm_mday = pDateTime->day +1;
|
|
tmbuf.tm_mon = pDateTime->month;
|
|
tmbuf.tm_year = pDateTime->year;
|
|
if (pDateTime->year < 40)
|
|
tmbuf.tm_year += 100; /* P8 uses 0-39 for 2000-2039 */
|
|
tmbuf.tm_wday = 0;
|
|
tmbuf.tm_yday = 0;
|
|
tmbuf.tm_isdst = -1; /* let it figure DST and time zone */
|
|
|
|
#if defined(HAVE_MKTIME)
|
|
when = mktime(&tmbuf);
|
|
#elif defined(HAVE_TIMELOCAL)
|
|
when = timelocal(&tmbuf);
|
|
#else
|
|
# error "need time converter"
|
|
#endif
|
|
|
|
if (when == (time_t) -1)
|
|
*pWhen = 0;
|
|
else
|
|
*pWhen = when;
|
|
}
|
|
|
|
/*
|
|
* Convert from GMT seconds since 1970 to local time in a NuDateTime struct.
|
|
*/
|
|
static void Nu_GMTSecondsToDateTime(const time_t* pWhen, NuDateTime *pDateTime)
|
|
{
|
|
struct tm* ptm;
|
|
|
|
Assert(pWhen != NULL);
|
|
Assert(pDateTime != NULL);
|
|
|
|
#if defined(HAVE_LOCALTIME_R) && defined(USE_REENTRANT_CALLS)
|
|
struct tm res;
|
|
ptm = localtime_r(pWhen, &res);
|
|
#else
|
|
/* NOTE: not thread-safe */
|
|
ptm = localtime(pWhen);
|
|
#endif
|
|
pDateTime->second = ptm->tm_sec;
|
|
pDateTime->minute = ptm->tm_min;
|
|
pDateTime->hour = ptm->tm_hour;
|
|
pDateTime->day = ptm->tm_mday -1;
|
|
pDateTime->month = ptm->tm_mon;
|
|
pDateTime->year = ptm->tm_year;
|
|
pDateTime->extra = 0;
|
|
pDateTime->weekDay = ptm->tm_wday +1;
|
|
}
|
|
#endif
|
|
|
|
|
|
/*
|
|
* Fill in the current time.
|
|
*/
|
|
void Nu_SetCurrentDateTime(NuDateTime* pDateTime)
|
|
{
|
|
Assert(pDateTime != NULL);
|
|
|
|
#if defined(UNIX_LIKE) || defined(WINDOWS_LIKE)
|
|
{
|
|
time_t now = time(NULL);
|
|
Nu_GMTSecondsToDateTime(&now, pDateTime);
|
|
}
|
|
#else
|
|
#error "Port this"
|
|
#endif
|
|
}
|
|
|
|
|
|
/*
|
|
* Returns "true" if "pWhen1" is older than "pWhen2". Returns false if
|
|
* "pWhen1" is the same age or newer than "pWhen2".
|
|
*
|
|
* On systems with mktime, it would be straightforward to convert the dates
|
|
* to time in seconds, and compare them that way. However, I don't want
|
|
* to rely on that function too heavily, so we just compare fields.
|
|
*/
|
|
Boolean Nu_IsOlder(const NuDateTime* pWhen1, const NuDateTime* pWhen2)
|
|
{
|
|
long result, year1, year2;
|
|
|
|
/* adjust for P8 ShrinkIt Y2K problem */
|
|
year1 = pWhen1->year;
|
|
if (year1 < 40)
|
|
year1 += 100;
|
|
year2 = pWhen2->year;
|
|
if (year2 < 40)
|
|
year2 += 100;
|
|
|
|
result = year1 - year2;
|
|
if (!result)
|
|
result = pWhen1->month - pWhen2->month;
|
|
if (!result)
|
|
result = pWhen1->day - pWhen2->day;
|
|
if (!result)
|
|
result = pWhen1->hour - pWhen2->hour;
|
|
if (!result)
|
|
result = pWhen1->minute - pWhen2->minute;
|
|
if (!result)
|
|
result = pWhen1->second - pWhen2->second;
|
|
|
|
if (result < 0)
|
|
return true;
|
|
return false;
|
|
}
|
|
|
|
|
|
/*
|
|
* ===========================================================================
|
|
* Get/set file info
|
|
* ===========================================================================
|
|
*/
|
|
|
|
/*
|
|
* System-independent (mostly) file info struct.
|
|
*/
|
|
typedef struct NuFileInfo {
|
|
Boolean isValid; /* init to "false", set "true" after we get data */
|
|
|
|
Boolean isRegularFile; /* is this a regular file? */
|
|
Boolean isDirectory; /* is this a directory? */
|
|
Boolean isForked; /* does file have a non-empty resource fork? */
|
|
|
|
uint32_t dataEof;
|
|
|
|
NuDateTime modWhen;
|
|
mode_t unixMode; /* UNIX-style permissions */
|
|
} NuFileInfo;
|
|
|
|
#define kDefaultFileType 0 /* "NON" */
|
|
#define kDefaultAuxType 0 /* $0000 */
|
|
|
|
|
|
/*
|
|
* Determine whether the record has both data and resource forks.
|
|
*/
|
|
static Boolean Nu_IsForkedFile(NuArchive* pArchive, const NuRecord* pRecord)
|
|
{
|
|
const NuThread* pThread;
|
|
NuThreadID threadID;
|
|
Boolean gotData, gotRsrc;
|
|
int i;
|
|
|
|
gotData = gotRsrc = false;
|
|
|
|
for (i = 0; i < (int)pRecord->recTotalThreads; i++) {
|
|
pThread = Nu_GetThread(pRecord, i);
|
|
Assert(pThread != NULL);
|
|
|
|
threadID = NuMakeThreadID(pThread->thThreadClass,pThread->thThreadKind);
|
|
if (threadID == kNuThreadIDDataFork)
|
|
gotData = true;
|
|
else if (threadID == kNuThreadIDRsrcFork)
|
|
gotRsrc = true;
|
|
}
|
|
|
|
if (gotData && gotRsrc)
|
|
return true;
|
|
else
|
|
return false;
|
|
}
|
|
|
|
|
|
#if defined(MAC_LIKE)
|
|
# if defined(HAS_RESOURCE_FORKS)
|
|
/*
|
|
* String to append to the filename to access the resource fork.
|
|
*
|
|
* This appears to be the correct way to access the resource fork, since
|
|
* at least OS X 10.1. Up until 10.7 ("Lion", July 2011), you could also
|
|
* access the fork with "/rsrc".
|
|
*/
|
|
static const char kMacRsrcPath[] = "/..namedfork/rsrc";
|
|
|
|
/*
|
|
* Generates the resource fork pathname from the file path.
|
|
*
|
|
* The caller must free the string returned.
|
|
*/
|
|
static UNICHAR* GetResourcePath(const UNICHAR* pathnameUNI)
|
|
{
|
|
Assert(pathnameUNI != NULL);
|
|
|
|
// sizeof(kMacRsrcPath) includes the string and the terminating null byte
|
|
const size_t bufLen =
|
|
strlen(pathnameUNI) * sizeof(UNICHAR) + sizeof(kMacRsrcPath);
|
|
char* buf;
|
|
|
|
buf = (char*) malloc(bufLen);
|
|
snprintf(buf, bufLen, "%s%s", pathnameUNI, kMacRsrcPath);
|
|
return buf;
|
|
}
|
|
# endif /*HAS_RESOURCE_FORKS*/
|
|
|
|
/*
|
|
* Due to historical reasons, the XATTR_FINDERINFO_NAME (defined to be
|
|
* ``com.apple.FinderInfo'') extended attribute must be 32 bytes; see the
|
|
* ATTR_CMN_FNDRINFO section in getattrlist(2).
|
|
*
|
|
* The FinderInfo block is the concatenation of a FileInfo structure
|
|
* and an ExtendedFileInfo (or ExtendedFolderInfo) structure -- see
|
|
* ATTR_CMN_FNDRINFO in getattrlist(2).
|
|
*
|
|
* All we're really interested in is the file type and creator code,
|
|
* which are stored big-endian in the first 8 bytes.
|
|
*/
|
|
static const int kFinderInfoSize = 32;
|
|
|
|
/*
|
|
* Set the file type and creator type.
|
|
*/
|
|
static NuError Nu_SetFinderInfo(NuArchive* pArchive, const NuRecord* pRecord,
|
|
const UNICHAR* pathnameUNI)
|
|
{
|
|
uint8_t fiBuf[kFinderInfoSize];
|
|
|
|
size_t actual = getxattr(pathnameUNI, XATTR_FINDERINFO_NAME,
|
|
fiBuf, sizeof(fiBuf), 0, 0);
|
|
if (actual == (size_t) -1 && errno == ENOATTR) {
|
|
// doesn't yet have Finder info
|
|
memset(fiBuf, 0, sizeof(fiBuf));
|
|
} else if (actual != kFinderInfoSize) {
|
|
Nu_ReportError(NU_BLOB, errno,
|
|
"Finder info on '%s' returned %d", pathnameUNI, (int) actual);
|
|
return kNuErrFile;
|
|
}
|
|
|
|
/* build type and creator from 8-bit type and 16-bit aux type */
|
|
uint32_t fileType, creator;
|
|
fileType = ('p' << 24) | ((pRecord->recFileType & 0xff) << 16) |
|
|
(pRecord->recExtraType & 0xffff);
|
|
creator = 'pdos';
|
|
|
|
fiBuf[0] = fileType >> 24;
|
|
fiBuf[1] = fileType >> 16;
|
|
fiBuf[2] = fileType >> 8;
|
|
fiBuf[3] = fileType;
|
|
fiBuf[4] = creator >> 24;
|
|
fiBuf[5] = creator >> 16;
|
|
fiBuf[6] = creator >> 8;
|
|
fiBuf[7] = creator;
|
|
|
|
if (setxattr(pathnameUNI, XATTR_FINDERINFO_NAME, fiBuf, sizeof(fiBuf),
|
|
0, 0) != 0)
|
|
{
|
|
Nu_ReportError(NU_BLOB, errno,
|
|
"Unable to set Finder info on '%s'", pathnameUNI);
|
|
return kNuErrFile;
|
|
}
|
|
|
|
return kNuErrNone;
|
|
}
|
|
#endif /*MAC_LIKE*/
|
|
|
|
|
|
/*
|
|
* Get the file info into a NuFileInfo struct. Fields which are
|
|
* inappropriate for the current system are set to default values.
|
|
*/
|
|
static NuError Nu_GetFileInfo(NuArchive* pArchive, const UNICHAR* pathnameUNI,
|
|
NuFileInfo* pFileInfo)
|
|
{
|
|
NuError err = kNuErrNone;
|
|
Assert(pArchive != NULL);
|
|
Assert(pathnameUNI != NULL);
|
|
Assert(pFileInfo != NULL);
|
|
|
|
pFileInfo->isValid = false;
|
|
|
|
#if defined(UNIX_LIKE) || defined(WINDOWS_LIKE)
|
|
{
|
|
struct stat sbuf;
|
|
int cc;
|
|
|
|
cc = stat(pathnameUNI, &sbuf);
|
|
if (cc) {
|
|
if (errno == ENOENT)
|
|
err = kNuErrFileNotFound;
|
|
else
|
|
err = kNuErrFileStat;
|
|
goto bail;
|
|
}
|
|
|
|
pFileInfo->isRegularFile = false;
|
|
if (S_ISREG(sbuf.st_mode))
|
|
pFileInfo->isRegularFile = true;
|
|
pFileInfo->isDirectory = false;
|
|
if (S_ISDIR(sbuf.st_mode))
|
|
pFileInfo->isDirectory = true;
|
|
|
|
/* BUG: should check for 32-bit overflow from 64-bit off_t */
|
|
pFileInfo->dataEof = sbuf.st_size;
|
|
pFileInfo->isForked = false;
|
|
# if defined(MAC_LIKE) && defined(HAS_RESOURCE_FORKS)
|
|
if (!pFileInfo->isDirectory) {
|
|
/*
|
|
* Check for the presence of a resource fork. You can check
|
|
* these from a terminal with "ls -l@" -- look for the
|
|
* "com.apple.ResourceFork" attribute.
|
|
*
|
|
* We can either use getxattr() and check for the presence of
|
|
* the attribute, or get the file length with stat(). I
|
|
* don't know if xattr has always worked with resource forks,
|
|
* so we'll stick with stat for now.
|
|
*/
|
|
UNICHAR* rsrcPath = GetResourcePath(pathnameUNI);
|
|
|
|
struct stat res_sbuf;
|
|
|
|
if (stat(rsrcPath, &res_sbuf) == 0) {
|
|
pFileInfo->isForked = (res_sbuf.st_size != 0);
|
|
}
|
|
|
|
free(rsrcPath);
|
|
}
|
|
# endif
|
|
Nu_GMTSecondsToDateTime(&sbuf.st_mtime, &pFileInfo->modWhen);
|
|
pFileInfo->unixMode = sbuf.st_mode;
|
|
pFileInfo->isValid = true;
|
|
}
|
|
#else
|
|
#error "Port this"
|
|
#endif
|
|
|
|
bail:
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* Determine whether a specific fork in the file exists.
|
|
*
|
|
* On systems that don't support forked files, the "checkRsrcFork" argument
|
|
* is ignored. If forked files are supported, and we are extracting a
|
|
* file with data and resource forks, we only claim it exists if it has
|
|
* nonzero length.
|
|
*/
|
|
static NuError Nu_FileForkExists(NuArchive* pArchive,
|
|
const UNICHAR* pathnameUNI, Boolean isForkedFile, Boolean checkRsrcFork,
|
|
Boolean* pExists, NuFileInfo* pFileInfo)
|
|
{
|
|
NuError err = kNuErrNone;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(pathnameUNI != NULL);
|
|
Assert(checkRsrcFork == true || checkRsrcFork == false);
|
|
Assert(pExists != NULL);
|
|
Assert(pFileInfo != NULL);
|
|
|
|
#if defined(UNIX_LIKE) || defined(WINDOWS_LIKE)
|
|
# if !defined(MAC_LIKE)
|
|
/*
|
|
* On Unix and Windows we ignore "isForkedFile" and "checkRsrcFork".
|
|
* The file must not exist at all.
|
|
*/
|
|
Assert(pArchive->lastFileCreatedUNI == NULL);
|
|
# endif
|
|
|
|
*pExists = true;
|
|
err = Nu_GetFileInfo(pArchive, pathnameUNI, pFileInfo);
|
|
if (err == kNuErrFileNotFound) {
|
|
err = kNuErrNone;
|
|
*pExists = false;
|
|
}
|
|
|
|
# if defined(MAC_LIKE)
|
|
/*
|
|
* On Mac OS X, we'll use the resource fork, but we may not want to
|
|
* overwrite existing data.
|
|
*/
|
|
if (*pExists && checkRsrcFork) {
|
|
*pExists = pFileInfo->isForked;
|
|
}
|
|
# endif
|
|
|
|
#elif defined(__ORCAC__)
|
|
/*
|
|
* If the file doesn't exist, great. If it does, and "lastFileCreated"
|
|
* matches up with this one, then we know that it exists because we
|
|
* created it.
|
|
*
|
|
* This is great unless the record has two data forks or something
|
|
* equally dopey, so we check to be sure that the fork we want to
|
|
* put the data into is currently empty.
|
|
*
|
|
* It is possible, though asinine, for a Mac OS or GS/OS extraction
|
|
* program to put the data and resource forks of a record into
|
|
* separate files, so we can't just assume that because we wrote
|
|
* the data fork to file A it is okay for file B to exist. That's
|
|
* why we compare the pathname instead of just remembering that
|
|
* we already created a file for this record.
|
|
*/
|
|
#error "Finish me"
|
|
|
|
#else
|
|
#error "Port this"
|
|
#endif
|
|
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* Set the dates on a file according to what's in the record.
|
|
*/
|
|
static NuError Nu_SetFileDates(NuArchive* pArchive, const NuRecord* pRecord,
|
|
const UNICHAR* pathnameUNI)
|
|
{
|
|
NuError err = kNuErrNone;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(pRecord != NULL);
|
|
Assert(pathnameUNI != NULL);
|
|
|
|
#if defined(UNIX_LIKE) || defined(WINDOWS_LIKE)
|
|
{
|
|
struct utimbuf utbuf;
|
|
|
|
/* ignore create time, and set access time equal to mod time */
|
|
Nu_DateTimeToGMTSeconds(&pRecord->recModWhen, &utbuf.modtime);
|
|
utbuf.actime = utbuf.modtime;
|
|
|
|
/* only do it if the NuDateTime was valid */
|
|
if (utbuf.modtime) {
|
|
if (utime(pathnameUNI, &utbuf) < 0) {
|
|
Nu_ReportError(NU_BLOB, errno,
|
|
"Unable to set time stamp on '%s'", pathnameUNI);
|
|
err = kNuErrFileSetDate;
|
|
goto bail;
|
|
}
|
|
}
|
|
}
|
|
|
|
#else
|
|
#error "Port this"
|
|
#endif
|
|
|
|
bail:
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* Returns "true" if the record is locked (in the ProDOS sense).
|
|
*
|
|
* Bits 31-8 reserved, must be zero
|
|
* Bit 7 (D) 1 = destroy enabled
|
|
* Bit 6 (R) 1 = rename enabled
|
|
* Bit 5 (B) 1 = file needs to be backed up
|
|
* Bits 4-3 reserved, must be zero
|
|
* Bit 2 (I) 1 = file is invisible
|
|
* Bit 1 (W) 1 = write enabled
|
|
* Bit 0 (R) 1 = read enabled
|
|
*
|
|
* A "locked" file would be 00?00001, "unlocked" 11?00011, with many
|
|
* possible variations. For our purposes, we treat all files as unlocked
|
|
* unless they match the classic "locked" bit pattern.
|
|
*/
|
|
static Boolean Nu_IsRecordLocked(const NuRecord* pRecord)
|
|
{
|
|
if (pRecord->recAccess == 0x21L || pRecord->recAccess == 0x01L)
|
|
return true;
|
|
else
|
|
return false;
|
|
}
|
|
|
|
/*
|
|
* Set the file access permissions based on what's in the record.
|
|
*
|
|
* This assumes that the file is currently writable, so we only need
|
|
* to do something if the original file was "locked".
|
|
*/
|
|
static NuError Nu_SetFileAccess(NuArchive* pArchive, const NuRecord* pRecord,
|
|
const UNICHAR* pathnameUNI)
|
|
{
|
|
NuError err = kNuErrNone;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(pRecord != NULL);
|
|
Assert(pathnameUNI != NULL);
|
|
|
|
#if defined(UNIX_LIKE) || defined(WINDOWS_LIKE)
|
|
/* only need to do something if the file was "locked" */
|
|
if (Nu_IsRecordLocked(pRecord)) {
|
|
mode_t mask;
|
|
|
|
/* set it to 444, modified by umask */
|
|
mask = umask(0);
|
|
umask(mask);
|
|
//DBUG(("+++ chmod '%s' %03o (mask=%03o)\n", pathname,
|
|
// (S_IRUSR | S_IRGRP | S_IROTH) & ~mask, mask));
|
|
if (chmod(pathnameUNI, (S_IRUSR | S_IRGRP | S_IROTH) & ~mask) < 0) {
|
|
Nu_ReportError(NU_BLOB, errno,
|
|
"unable to set access for '%s' to %03o", pathnameUNI,
|
|
(int) mask);
|
|
err = kNuErrFileSetAccess;
|
|
goto bail;
|
|
}
|
|
}
|
|
|
|
#else
|
|
#error "Port this"
|
|
#endif
|
|
|
|
bail:
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* ===========================================================================
|
|
* Create/open an output file
|
|
* ===========================================================================
|
|
*/
|
|
|
|
/*
|
|
* Prepare an existing file for writing.
|
|
*
|
|
* Generally this just involves ensuring that the file is writable. If
|
|
* this is a convenient place to truncate it, we should do that too.
|
|
*
|
|
* 20150103: we don't seem to be doing the truncation here, so prepRsrc
|
|
* is unused.
|
|
*/
|
|
static NuError Nu_PrepareForWriting(NuArchive* pArchive,
|
|
const UNICHAR* pathnameUNI, Boolean prepRsrc, NuFileInfo* pFileInfo)
|
|
{
|
|
NuError err = kNuErrNone;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(pathnameUNI != NULL);
|
|
Assert(prepRsrc == true || prepRsrc == false);
|
|
Assert(pFileInfo != NULL);
|
|
|
|
Assert(pFileInfo->isValid == true);
|
|
|
|
/* don't go playing with directories, pipes, etc */
|
|
if (pFileInfo->isRegularFile != true)
|
|
return kNuErrNotRegularFile;
|
|
|
|
#if defined(UNIX_LIKE) || defined(WINDOWS_LIKE)
|
|
if (!(pFileInfo->unixMode & S_IWUSR)) {
|
|
/* make it writable by owner, plus whatever it was before */
|
|
if (chmod(pathnameUNI, S_IWUSR | pFileInfo->unixMode) < 0) {
|
|
Nu_ReportError(NU_BLOB, errno,
|
|
"unable to set access for '%s'", pathnameUNI);
|
|
err = kNuErrFileSetAccess;
|
|
goto bail;
|
|
}
|
|
}
|
|
|
|
return kNuErrNone;
|
|
|
|
#else
|
|
#error "Port this"
|
|
#endif
|
|
|
|
bail:
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* Invoke the system-dependent directory creation function.
|
|
*/
|
|
static NuError Nu_Mkdir(NuArchive* pArchive, const char* dir)
|
|
{
|
|
NuError err = kNuErrNone;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(dir != NULL);
|
|
|
|
#if defined(UNIX_LIKE)
|
|
if (mkdir(dir, S_IRWXU | S_IRGRP|S_IXGRP | S_IROTH|S_IXOTH) < 0) {
|
|
err = errno ? errno : kNuErrDirCreate;
|
|
Nu_ReportError(NU_BLOB, err, "Unable to create dir '%s'", dir);
|
|
goto bail;
|
|
}
|
|
|
|
#elif defined(WINDOWS_LIKE)
|
|
if (mkdir(dir) < 0) {
|
|
err = errno ? errno : kNuErrDirCreate;
|
|
Nu_ReportError(NU_BLOB, err, "Unable to create dir '%s'", dir);
|
|
goto bail;
|
|
}
|
|
|
|
#else
|
|
#error "Port this"
|
|
#endif
|
|
|
|
bail:
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* Create a single subdirectory if it doesn't exist. If the next-highest
|
|
* subdirectory level doesn't exist either, cut down the pathname and
|
|
* recurse.
|
|
*/
|
|
static NuError Nu_CreateSubdirIFN(NuArchive* pArchive,
|
|
const UNICHAR* pathStartUNI, const char* pathEnd, char fssep)
|
|
{
|
|
NuError err = kNuErrNone;
|
|
NuFileInfo fileInfo;
|
|
char* tmpBuf = NULL;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(pathStartUNI != NULL);
|
|
Assert(pathEnd != NULL);
|
|
Assert(fssep != '\0');
|
|
|
|
/* pathStart might have whole path, but we only want up to "pathEnd" */
|
|
tmpBuf = strdup(pathStartUNI);
|
|
tmpBuf[pathEnd - pathStartUNI +1] = '\0';
|
|
|
|
err = Nu_GetFileInfo(pArchive, tmpBuf, &fileInfo);
|
|
if (err == kNuErrFileNotFound) {
|
|
/* dir doesn't exist; move up a level and check parent */
|
|
pathEnd = strrchr(tmpBuf, fssep);
|
|
if (pathEnd != NULL) {
|
|
pathEnd--;
|
|
Assert(pathEnd >= tmpBuf);
|
|
err = Nu_CreateSubdirIFN(pArchive, tmpBuf, pathEnd, fssep);
|
|
BailError(err);
|
|
}
|
|
|
|
/* parent is taken care of; create this one */
|
|
err = Nu_Mkdir(pArchive, tmpBuf);
|
|
BailError(err);
|
|
} else if (err != kNuErrNone) {
|
|
goto bail;
|
|
} else {
|
|
/* file does exist, make sure it's a directory */
|
|
Assert(fileInfo.isValid == true);
|
|
if (!fileInfo.isDirectory) {
|
|
err = kNuErrNotDir;
|
|
Nu_ReportError(NU_BLOB, err, "Unable to create path '%s'", tmpBuf);
|
|
goto bail;
|
|
}
|
|
}
|
|
|
|
bail:
|
|
Nu_Free(pArchive, tmpBuf);
|
|
return err;
|
|
}
|
|
|
|
/*
|
|
* Create subdirectories, if needed. The paths leading up to the filename
|
|
* in "pathname" will be created.
|
|
*
|
|
* If "pathname" is just a filename, or the set of directories matches
|
|
* the last directory we created, we don't do anything.
|
|
*/
|
|
static NuError Nu_CreatePathIFN(NuArchive* pArchive, const UNICHAR* pathnameUNI,
|
|
UNICHAR fssep)
|
|
{
|
|
NuError err = kNuErrNone;
|
|
const char* pathStart;
|
|
const char* pathEnd;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(pathnameUNI != NULL);
|
|
Assert(fssep != '\0');
|
|
|
|
pathStart = pathnameUNI;
|
|
|
|
#if !defined(MAC_LIKE) /* On the Mac, if it's a full path, treat it like one */
|
|
// 20150103: not sure what use case this is for
|
|
if (pathnameUNI[0] == fssep)
|
|
pathStart++;
|
|
#endif
|
|
|
|
/* NOTE: not expecting names like "foo/bar/ack/", with terminating fssep */
|
|
pathEnd = strrchr(pathStart, fssep);
|
|
if (pathEnd == NULL) {
|
|
/* no subdirectory components found */
|
|
goto bail;
|
|
}
|
|
pathEnd--;
|
|
|
|
Assert(pathEnd >= pathStart);
|
|
if (pathEnd - pathStart < 0)
|
|
goto bail;
|
|
|
|
/*
|
|
* On some filesystems, strncasecmp would be appropriate here. However,
|
|
* this is meant solely as an optimization to avoid extra stat() calls,
|
|
* so we want to use the most restrictive case.
|
|
*/
|
|
if (pArchive->lastDirCreatedUNI &&
|
|
strncmp(pathStart, pArchive->lastDirCreatedUNI,
|
|
pathEnd - pathStart +1) == 0)
|
|
{
|
|
/* we created this one recently, don't do it again */
|
|
goto bail;
|
|
}
|
|
|
|
/*
|
|
* Test to determine which directories exist. The most likely case
|
|
* is that some or all of the components have already been created,
|
|
* so we start with the last one and work backward.
|
|
*/
|
|
err = Nu_CreateSubdirIFN(pArchive, pathStart, pathEnd, fssep);
|
|
BailError(err);
|
|
|
|
bail:
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* Open the file for writing, possibly truncating it.
|
|
*/
|
|
static NuError Nu_OpenFileForWrite(NuArchive* pArchive,
|
|
const UNICHAR* pathnameUNI, Boolean openRsrc, FILE** pFp)
|
|
{
|
|
#if defined(MAC_LIKE) && defined(HAS_RESOURCE_FORKS)
|
|
if (openRsrc) {
|
|
UNICHAR* rsrcPath = GetResourcePath(pathnameUNI);
|
|
*pFp = fopen(rsrcPath, kNuFileOpenWriteTrunc);
|
|
free(rsrcPath);
|
|
} else {
|
|
*pFp = fopen(pathnameUNI, kNuFileOpenWriteTrunc);
|
|
}
|
|
#else
|
|
*pFp = fopen(pathnameUNI, kNuFileOpenWriteTrunc);
|
|
#endif
|
|
if (*pFp == NULL)
|
|
return errno ? errno : -1;
|
|
return kNuErrNone;
|
|
}
|
|
|
|
|
|
/*
|
|
* Open an output file and prepare it for writing.
|
|
*
|
|
* There are a number of things to take into consideration, including
|
|
* deal with "file exists" conditions, handling Mac/IIgs file types,
|
|
* coping with resource forks on extended files, and handling the
|
|
* "freshen" option that requires us to only update files that are
|
|
* older than what we have.
|
|
*/
|
|
NuError Nu_OpenOutputFile(NuArchive* pArchive, const NuRecord* pRecord,
|
|
const NuThread* pThread, const UNICHAR* newPathnameUNI, UNICHAR newFssep,
|
|
FILE** pFp)
|
|
{
|
|
NuError err = kNuErrNone;
|
|
Boolean exists, isForkedFile, extractingRsrc = false;
|
|
NuFileInfo fileInfo;
|
|
NuErrorStatus errorStatus;
|
|
NuResult result;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(pRecord != NULL);
|
|
Assert(pThread != NULL);
|
|
Assert(newPathnameUNI != NULL);
|
|
Assert(pFp != NULL);
|
|
|
|
/* set up some defaults, in case something goes wrong */
|
|
errorStatus.operation = kNuOpExtract;
|
|
errorStatus.err = kNuErrInternal;
|
|
errorStatus.sysErr = 0;
|
|
errorStatus.message = NULL;
|
|
errorStatus.pRecord = pRecord;
|
|
errorStatus.pathnameUNI = newPathnameUNI;
|
|
errorStatus.origPathname = NULL;
|
|
errorStatus.filenameSeparator = newFssep;
|
|
/*errorStatus.origArchiveTouched = false;*/
|
|
errorStatus.canAbort = true;
|
|
errorStatus.canRetry = true;
|
|
errorStatus.canIgnore = false;
|
|
errorStatus.canSkip = true;
|
|
errorStatus.canRename = true;
|
|
errorStatus.canOverwrite = true;
|
|
|
|
/* decide if this is a forked file (i.e. has *both* forks) */
|
|
isForkedFile = Nu_IsForkedFile(pArchive, pRecord);
|
|
|
|
/* decide if we're extracting a resource fork */
|
|
if (NuMakeThreadID(pThread->thThreadClass, pThread->thThreadKind) ==
|
|
kNuThreadIDRsrcFork)
|
|
{
|
|
extractingRsrc = true;
|
|
}
|
|
|
|
/*
|
|
* Determine whether the file and fork already exists. If the file
|
|
* is one we just created, and the fork we want to write to is
|
|
* empty, this will *not* set "exists".
|
|
*/
|
|
fileInfo.isValid = false;
|
|
err = Nu_FileForkExists(pArchive, newPathnameUNI, isForkedFile,
|
|
extractingRsrc, &exists, &fileInfo);
|
|
BailError(err);
|
|
|
|
if (exists) {
|
|
Assert(fileInfo.isValid == true);
|
|
|
|
/*
|
|
* The file exists when it shouldn't. Decide what to do, based
|
|
* on the options configured by the application.
|
|
*/
|
|
|
|
/*
|
|
* Start by checking to see if we're willing to overwrite older files.
|
|
* If not, see if the application wants to rename the file, or force
|
|
* the overwrite. Most likely they'll just want to skip it.
|
|
*/
|
|
if ((pArchive->valOnlyUpdateOlder) &&
|
|
!Nu_IsOlder(&fileInfo.modWhen, &pRecord->recModWhen))
|
|
{
|
|
if (pArchive->errorHandlerFunc != NULL) {
|
|
errorStatus.err = kNuErrNotNewer;
|
|
result = (*pArchive->errorHandlerFunc)(pArchive, &errorStatus);
|
|
|
|
switch (result) {
|
|
case kNuAbort:
|
|
err = kNuErrAborted;
|
|
goto bail;
|
|
case kNuRetry:
|
|
case kNuRename:
|
|
err = kNuErrRename;
|
|
goto bail;
|
|
case kNuSkip:
|
|
err = kNuErrSkipped;
|
|
goto bail;
|
|
case kNuOverwrite:
|
|
break; /* fall back into main code */
|
|
case kNuIgnore:
|
|
default:
|
|
err = kNuErrSyntax;
|
|
Nu_ReportError(NU_BLOB, err,
|
|
"Wasn't expecting result %d here", result);
|
|
goto bail;
|
|
}
|
|
} else {
|
|
err = kNuErrNotNewer;
|
|
goto bail;
|
|
}
|
|
}
|
|
|
|
/* If they "might" allow overwrites, and they have an error-handling
|
|
* callback defined, call that to find out what they want to do
|
|
* here. Options include skipping the file, overwriting the file,
|
|
* and extracting to a different file.
|
|
*/
|
|
if (pArchive->valHandleExisting == kNuMaybeOverwrite) {
|
|
if (pArchive->errorHandlerFunc != NULL) {
|
|
errorStatus.err = kNuErrFileExists;
|
|
result = (*pArchive->errorHandlerFunc)(pArchive, &errorStatus);
|
|
|
|
switch (result) {
|
|
case kNuAbort:
|
|
err = kNuErrAborted;
|
|
goto bail;
|
|
case kNuRetry:
|
|
case kNuRename:
|
|
err = kNuErrRename;
|
|
goto bail;
|
|
case kNuSkip:
|
|
err = kNuErrSkipped;
|
|
goto bail;
|
|
case kNuOverwrite:
|
|
break; /* fall back into main code */
|
|
case kNuIgnore:
|
|
default:
|
|
err = kNuErrSyntax;
|
|
Nu_ReportError(NU_BLOB, err,
|
|
"Wasn't expecting result %d here", result);
|
|
goto bail;
|
|
}
|
|
} else {
|
|
/* no error handler, return an error to the caller */
|
|
err = kNuErrFileExists;
|
|
goto bail;
|
|
}
|
|
} else if (pArchive->valHandleExisting == kNuNeverOverwrite) {
|
|
err = kNuErrSkipped;
|
|
goto bail;
|
|
}
|
|
} else {
|
|
/*
|
|
* The file doesn't exist. If we're doing a "freshen" from the
|
|
* archive, we don't want to create a new file, so we return an
|
|
* error to the user instead. However, we give the application
|
|
* a chance to straighten things out. Most likely they'll just
|
|
* return kNuSkip.
|
|
*/
|
|
if (pArchive->valHandleExisting == kNuMustOverwrite) {
|
|
DBUG(("+++ can't freshen nonexistent file '%s'\n", newPathnameUNI));
|
|
if (pArchive->errorHandlerFunc != NULL) {
|
|
errorStatus.err = kNuErrDuplicateNotFound;
|
|
|
|
/* give them a chance to rename */
|
|
result = (*pArchive->errorHandlerFunc)(pArchive, &errorStatus);
|
|
switch (result) {
|
|
case kNuAbort:
|
|
err = kNuErrAborted;
|
|
goto bail;
|
|
case kNuRetry:
|
|
case kNuRename:
|
|
err = kNuErrRename;
|
|
goto bail;
|
|
case kNuSkip:
|
|
err = kNuErrSkipped;
|
|
goto bail;
|
|
case kNuOverwrite:
|
|
break; /* fall back into main code */
|
|
case kNuIgnore:
|
|
default:
|
|
err = kNuErrSyntax;
|
|
Nu_ReportError(NU_BLOB, err,
|
|
"Wasn't expecting result %d here", result);
|
|
goto bail;
|
|
}
|
|
} else {
|
|
/* no error handler, return an error to the caller */
|
|
err = kNuErrDuplicateNotFound;
|
|
goto bail;
|
|
}
|
|
}
|
|
}
|
|
|
|
Assert(err == kNuErrNone);
|
|
|
|
/*
|
|
* After the above, if the file exists then we need to prepare it for
|
|
* writing. On some systems -- notably those with forked files -- it
|
|
* may be easiest to delete the entire file and start over. On
|
|
* simpler systems, an (optional) chmod followed by an open that
|
|
* truncates the file should be sufficient.
|
|
*
|
|
* If the file didn't exist, we need to be sure that the path leading
|
|
* up to its eventual location exists. This might require creating
|
|
* several directories. We distinguish the case of "file isn't there"
|
|
* from "file is there but fork isn't" by seeing if we were able to
|
|
* get valid file info.
|
|
*/
|
|
if (exists) {
|
|
Assert(fileInfo.isValid == true);
|
|
err = Nu_PrepareForWriting(pArchive, newPathnameUNI, extractingRsrc,
|
|
&fileInfo);
|
|
BailError(err);
|
|
} else if (!fileInfo.isValid) {
|
|
err = Nu_CreatePathIFN(pArchive, newPathnameUNI, newFssep);
|
|
BailError(err);
|
|
}
|
|
|
|
/*
|
|
* Open sesame.
|
|
*/
|
|
err = Nu_OpenFileForWrite(pArchive, newPathnameUNI, extractingRsrc, pFp);
|
|
BailError(err);
|
|
|
|
|
|
#if defined(HAS_RESOURCE_FORKS)
|
|
pArchive->lastFileCreatedUNI = newPathnameUNI;
|
|
#endif
|
|
|
|
bail:
|
|
if (err != kNuErrNone) {
|
|
if (err != kNuErrSkipped && err != kNuErrRename &&
|
|
err != kNuErrFileExists)
|
|
{
|
|
Nu_ReportError(NU_BLOB, err, "Unable to open '%s'%s",
|
|
newPathnameUNI, extractingRsrc ? " (rsrc fork)" : "");
|
|
}
|
|
}
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* Close the output file, adjusting the modification date and access
|
|
* permissions as needed.
|
|
*
|
|
* On GS/OS and Mac OS, we may need to set the file type here, depending on
|
|
* how much we managed to do when the file was first created. IIRC,
|
|
* the GS/OS Open call should allow setting the file type.
|
|
*
|
|
* BUG: on GS/OS, if we set the file access after writing the data fork,
|
|
* we may not be able to open the same file for writing the rsrc fork.
|
|
* We can't suppress setting the access permissions, because we don't know
|
|
* if the application will want to write both forks to the same file, or
|
|
* for that matter will want to write the resource fork at all. Looks
|
|
* like we will have to be smart enough to reset the access permissions
|
|
* when writing a rsrc fork to a file with just a data fork. This isn't
|
|
* quite right, but it's close enough.
|
|
*/
|
|
NuError Nu_CloseOutputFile(NuArchive* pArchive, const NuRecord* pRecord,
|
|
FILE* fp, const UNICHAR* pathnameUNI)
|
|
{
|
|
NuError err;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(pRecord != NULL);
|
|
Assert(fp != NULL);
|
|
|
|
fclose(fp);
|
|
|
|
err = Nu_SetFileDates(pArchive, pRecord, pathnameUNI);
|
|
BailError(err);
|
|
|
|
err = Nu_SetFileAccess(pArchive, pRecord, pathnameUNI);
|
|
BailError(err);
|
|
|
|
#if defined(MAC_LIKE)
|
|
/* could also do this earlier and pass the fd for fsetxattr */
|
|
err = Nu_SetFinderInfo(pArchive, pRecord, pathnameUNI);
|
|
BailError(err);
|
|
#endif
|
|
|
|
bail:
|
|
return kNuErrNone;
|
|
}
|
|
|
|
/*
|
|
* ===========================================================================
|
|
* Open an input file
|
|
* ===========================================================================
|
|
*/
|
|
|
|
/*
|
|
* Open the file for reading, in "binary" mode when necessary.
|
|
*/
|
|
static NuError Nu_OpenFileForRead(NuArchive* pArchive,
|
|
const UNICHAR* pathnameUNI, Boolean openRsrc, FILE** pFp)
|
|
{
|
|
*pFp = fopen(pathnameUNI, kNuFileOpenReadOnly);
|
|
if (*pFp == NULL)
|
|
return errno ? errno : -1;
|
|
return kNuErrNone;
|
|
}
|
|
|
|
|
|
/*
|
|
* Open an input file and prepare it for reading.
|
|
*
|
|
* If the file can't be found, we give the application an opportunity to
|
|
* skip the absent file, retry, or abort the whole thing.
|
|
*/
|
|
NuError Nu_OpenInputFile(NuArchive* pArchive, const UNICHAR* pathnameUNI,
|
|
Boolean openRsrc, FILE** pFp)
|
|
{
|
|
NuError err = kNuErrNone;
|
|
NuError openErr = kNuErrNone;
|
|
NuErrorStatus errorStatus;
|
|
NuResult result;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(pathnameUNI != NULL);
|
|
Assert(pFp != NULL);
|
|
|
|
#if defined(MAC_LIKE) && defined(HAS_RESOURCE_FORKS)
|
|
UNICHAR* rsrcPath = NULL;
|
|
if (openRsrc) {
|
|
rsrcPath = GetResourcePath(pathnameUNI);
|
|
pathnameUNI = rsrcPath;
|
|
}
|
|
#endif
|
|
|
|
retry:
|
|
/*
|
|
* Open sesame.
|
|
*/
|
|
err = Nu_OpenFileForRead(pArchive, pathnameUNI, openRsrc, pFp);
|
|
if (err == kNuErrNone) /* success! */
|
|
goto bail;
|
|
|
|
if (err == ENOENT)
|
|
openErr = kNuErrFileNotFound;
|
|
|
|
if (pArchive->errorHandlerFunc != NULL) {
|
|
errorStatus.operation = kNuOpAdd;
|
|
errorStatus.err = openErr;
|
|
errorStatus.sysErr = 0;
|
|
errorStatus.message = NULL;
|
|
errorStatus.pRecord = NULL;
|
|
errorStatus.pathnameUNI = pathnameUNI;
|
|
errorStatus.origPathname = NULL;
|
|
errorStatus.filenameSeparator = '\0';
|
|
/*errorStatus.origArchiveTouched = false;*/
|
|
errorStatus.canAbort = true;
|
|
errorStatus.canRetry = true;
|
|
errorStatus.canIgnore = false;
|
|
errorStatus.canSkip = true;
|
|
errorStatus.canRename = false;
|
|
errorStatus.canOverwrite = false;
|
|
|
|
DBUG(("--- invoking error handler function\n"));
|
|
result = (*pArchive->errorHandlerFunc)(pArchive, &errorStatus);
|
|
|
|
switch (result) {
|
|
case kNuAbort:
|
|
err = kNuErrAborted;
|
|
goto bail;
|
|
case kNuRetry:
|
|
goto retry;
|
|
case kNuSkip:
|
|
err = kNuErrSkipped;
|
|
goto bail;
|
|
case kNuRename:
|
|
case kNuOverwrite:
|
|
case kNuIgnore:
|
|
default:
|
|
err = kNuErrSyntax;
|
|
Nu_ReportError(NU_BLOB, err,
|
|
"Wasn't expecting result %d here", result);
|
|
goto bail;
|
|
}
|
|
} else {
|
|
DBUG(("+++ no error callback in OpenInputFile\n"));
|
|
}
|
|
|
|
bail:
|
|
if (err == kNuErrNone) {
|
|
Assert(*pFp != NULL);
|
|
} else {
|
|
if (err != kNuErrSkipped && err != kNuErrRename &&
|
|
err != kNuErrFileExists)
|
|
{
|
|
Nu_ReportError(NU_BLOB, err, "Unable to open '%s'%s",
|
|
pathnameUNI, openRsrc ? " (rsrc fork)" : "");
|
|
}
|
|
}
|
|
#if defined(MAC_LIKE) && defined(HAS_RESOURCE_FORKS)
|
|
free(rsrcPath);
|
|
#endif
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* ===========================================================================
|
|
* Delete and rename files
|
|
* ===========================================================================
|
|
*/
|
|
|
|
/*
|
|
* Delete a file.
|
|
*/
|
|
NuError Nu_DeleteFile(const UNICHAR* pathnameUNI)
|
|
{
|
|
#if defined(UNIX_LIKE) || defined(WINDOWS_LIKE)
|
|
int cc;
|
|
|
|
DBUG(("--- Deleting '%s'\n", pathnameUNI));
|
|
|
|
cc = unlink(pathnameUNI);
|
|
if (cc < 0)
|
|
return errno ? errno : -1;
|
|
else
|
|
return kNuErrNone;
|
|
#else
|
|
#error "Port this"
|
|
#endif
|
|
}
|
|
|
|
/*
|
|
* Rename a file from "fromPath" to "toPath".
|
|
*/
|
|
NuError Nu_RenameFile(const UNICHAR* fromPathUNI, const UNICHAR* toPathUNI)
|
|
{
|
|
#if defined(UNIX_LIKE) || defined(WINDOWS_LIKE)
|
|
int cc;
|
|
|
|
DBUG(("--- Renaming '%s' to '%s'\n", fromPathUNI, toPathUNI));
|
|
|
|
cc = rename(fromPathUNI, toPathUNI);
|
|
if (cc < 0)
|
|
return errno ? errno : -1;
|
|
else
|
|
return kNuErrNone;
|
|
#else
|
|
#error "Port this"
|
|
#endif
|
|
}
|
|
|
|
|
|
/*
|
|
* ===========================================================================
|
|
* NuError wrappers for std functions
|
|
* ===========================================================================
|
|
*/
|
|
|
|
/*
|
|
* Wrapper for ftell().
|
|
*/
|
|
NuError Nu_FTell(FILE* fp, long* pOffset)
|
|
{
|
|
Assert(fp != NULL);
|
|
Assert(pOffset != NULL);
|
|
|
|
errno = 0;
|
|
*pOffset = ftell(fp);
|
|
if (*pOffset < 0) {
|
|
Nu_ReportError(NU_NILBLOB, errno, "file ftell failed");
|
|
return errno ? errno : kNuErrFileSeek;
|
|
}
|
|
return kNuErrNone;
|
|
}
|
|
|
|
/*
|
|
* Wrapper for fseek().
|
|
*/
|
|
NuError Nu_FSeek(FILE* fp, long offset, int ptrname)
|
|
{
|
|
Assert(fp != NULL);
|
|
Assert(ptrname == SEEK_SET || ptrname == SEEK_CUR || ptrname == SEEK_END);
|
|
|
|
errno = 0;
|
|
if (fseek(fp, offset, ptrname) < 0) {
|
|
Nu_ReportError(NU_NILBLOB, errno,
|
|
"file fseek(%ld, %d) failed", offset, ptrname);
|
|
return errno ? errno : kNuErrFileSeek;
|
|
}
|
|
return kNuErrNone;
|
|
}
|
|
|
|
/*
|
|
* Wrapper for fread(). Note the arguments resemble read(2) rather than the
|
|
* slightly silly ones used by fread(3S).
|
|
*/
|
|
NuError Nu_FRead(FILE* fp, void* buf, size_t nbyte)
|
|
{
|
|
size_t result;
|
|
|
|
errno = 0;
|
|
result = fread(buf, nbyte, 1, fp);
|
|
if (result != 1)
|
|
return errno ? errno : kNuErrFileRead;
|
|
return kNuErrNone;
|
|
}
|
|
|
|
/*
|
|
* Wrapper for fwrite(). Note the arguments resemble write(2) rather than the
|
|
* slightly silly ones used by fwrite(3S).
|
|
*/
|
|
NuError Nu_FWrite(FILE* fp, const void* buf, size_t nbyte)
|
|
{
|
|
size_t result;
|
|
|
|
errno = 0;
|
|
result = fwrite(buf, nbyte, 1, fp);
|
|
if (result != 1)
|
|
return errno ? errno : kNuErrFileWrite;
|
|
return kNuErrNone;
|
|
}
|
|
|
|
/*
|
|
* ===========================================================================
|
|
* Misc functions
|
|
* ===========================================================================
|
|
*/
|
|
|
|
/*
|
|
* Copy a section from one file to another.
|
|
*/
|
|
NuError Nu_CopyFileSection(NuArchive* pArchive, FILE* dstFp, FILE* srcFp,
|
|
long length)
|
|
{
|
|
NuError err;
|
|
long readLen;
|
|
|
|
Assert(pArchive != NULL);
|
|
Assert(dstFp != NULL);
|
|
Assert(srcFp != NULL);
|
|
Assert(length >= 0); /* can be == 0, e.g. empty data fork from HFS */
|
|
|
|
/* nice big buffer, for speed... could use getc/putc for simplicity */
|
|
err = Nu_AllocCompressionBufferIFN(pArchive);
|
|
BailError(err);
|
|
|
|
DBUG(("+++ Copying %ld bytes\n", length));
|
|
|
|
while (length) {
|
|
readLen = length > kNuGenCompBufSize ? kNuGenCompBufSize : length;
|
|
|
|
err = Nu_FRead(srcFp, pArchive->compBuf, readLen);
|
|
if (err != kNuErrNone) {
|
|
Nu_ReportError(NU_BLOB, err,
|
|
"Nu_FRead failed while copying file section "
|
|
"(fp=0x%08lx, readLen=%ld, length=%ld, err=%d)\n",
|
|
(long) srcFp, readLen, length, err);
|
|
goto bail;
|
|
}
|
|
err = Nu_FWrite(dstFp, pArchive->compBuf, readLen);
|
|
BailError(err);
|
|
|
|
length -= readLen;
|
|
}
|
|
|
|
bail:
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* Find the length of an open file.
|
|
*
|
|
* On UNIX it would be easier to just call fstat(), but fseek is portable.
|
|
*
|
|
* Only useful for files < 2GB in size.
|
|
*
|
|
* (pArchive is only used for BailError message reporting, so it's okay
|
|
* to call here with a NULL pointer if the archive isn't open yet.)
|
|
*/
|
|
NuError Nu_GetFileLength(NuArchive* pArchive, FILE* fp, long* pLength)
|
|
{
|
|
NuError err;
|
|
long oldpos;
|
|
|
|
Assert(fp != NULL);
|
|
Assert(pLength != NULL);
|
|
|
|
err = Nu_FTell(fp, &oldpos);
|
|
BailError(err);
|
|
|
|
err = Nu_FSeek(fp, 0, SEEK_END);
|
|
BailError(err);
|
|
|
|
err = Nu_FTell(fp, pLength);
|
|
BailError(err);
|
|
|
|
err = Nu_FSeek(fp, oldpos, SEEK_SET);
|
|
BailError(err);
|
|
|
|
bail:
|
|
return err;
|
|
}
|
|
|
|
|
|
/*
|
|
* Truncate an open file. This differs from ftruncate() in that it takes
|
|
* a FILE* instead of an fd, and the length is a long instead of off_t.
|
|
*/
|
|
NuError Nu_TruncateOpenFile(FILE* fp, long length)
|
|
{
|
|
#if defined(HAVE_FTRUNCATE)
|
|
if (ftruncate(fileno(fp), length) < 0)
|
|
return errno ? errno : -1;
|
|
return kNuErrNone;
|
|
#elif defined(HAVE_CHSIZE)
|
|
if (chsize(fileno(fp), length) < 0)
|
|
return errno ? errno : -1;
|
|
return kNuErrNone;
|
|
#else
|
|
/* not fatal; return this to indicate that it's an unsupported operation */
|
|
return kNuErrInternal;
|
|
#endif
|
|
}
|
|
|