Here is a part of code in [unixWrite]. (os_unix.h)
Is not it written by mmap ?
#if SQLITE_MAX_MMAP_SIZE0
/* Deal with as much of this write request as possible by transfering
** data from the memory mapping using memcpy(). */
if( offsetpFile-mmapSize ){
if( offset+amt = pFile-mmapSize ){
int size = -1;
struct stat buf; /* Used to hold return values of fstat() */
if( !osFstat(pFile-h, buf) ){
size = (int)buf.st_size;
}
memcpy(((u8 *)(pFile-pMapRegion))[offset], pBuf, amt);
return SQLITE_OK;
}else{
int nCopy = (int)(pFile-mmapSize - offset);
memcpy(((u8 *)(pFile-pMapRegion))[offset], pBuf, nCopy);
pBuf = ((u8 *)pBuf)[nCopy];
amt -= nCopy;
offset += nCopy;
}
}
#endif
原始邮件
发件人:Richard [email protected]
收件人:SQLite mailing [email protected]
发送时间:2016年11月3日(周四) 17:49
主题:Re: [sqlite] Is msync necessary for SQLite?
On 11/3/16, sanhua.zh [email protected] wrote: I run in iOS. The mmap is
enable in recent version. So, once a mmap enable, data may lost after os crash
unless msync. But I can’t find anywhere in the source of SQLite call the
msync. Is it a bug? No. Mmap() is only used for reading, not writing. All
changes are implemented by calling write() and fsync(). -- D. Richard Hipp
[email protected] _______________________________________________ sqlite-users
mailing list [email protected]
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
_______________________________________________
sqlite-users mailing list
[email protected]
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
