From: Nguyen Thai Ngoc Duy Date: Sat, 21 Apr 2012 03:00:18 GMT Subject: Re: What's cooking in git.git (Apr 2012, #06; Sun, 15) Message-ID: <20120421030018.GA32687@do> In-Reply-To: On Thu, Apr 19, 2012 at 03:52:44PM +0200, Erik Faye-Lund wrote: > On Thu, Apr 19, 2012 at 3:48 PM, Johannes Sixt wrote: > > Am 4/19/2012 15:31, schrieb Erik Faye-Lund: > >> int main(int argc, const char *argv[]) > >> { > >>         int i, fd = open(__FILE__, O_RDONLY); > >>         for (i = 0; i < 2; ++i) { > >>                 char buf[11] = {0}; > >>                 mingw_pread(fd, buf, 10, 0); > >>                 printf("buf = '%10s'\n", buf); > >>         } > >>         return 0; > >> } > > > > The test is flawed. It shows only that pread can read twice the same file > > location. > > Oh, you are of course right. *slaps forehead* > > > But it must not update the file pointer, whereas the > > documentation of ReadFile says (quoting the paragraph that is relevant for > > us): > > > >  If hFile is not opened with FILE_FLAG_OVERLAPPED and lpOverlapped is > >  not NULL, the read operation starts at the offset that is specified in > >  the OVERLAPPED structure. ReadFile does not return until the read > >  operation is complete, and then the system updates the file pointer. > > Indeed. Correcting the test to first do pread and then read reveals > that the file pointer does indeed get updated. Grrr... > > > Hence, your mingw_pread() looks like a workable solution for our purposes, > > but is still not a 100% correct emulation of pread(). (Testing all this > > will have to wait for another 18 hours or so.) > > > > I don't know our use-cases, but I'd be a lot happier if I could find a > safe way to have it not update the file-pointer. Just reading it and > setting it back again would be racy. Replace pread() in index-pack to pread_weak() because we know we don't care about file offset in index-pack. Define pread_weak as pread normally. Windows port can provide its own pread_weak version, which can freely move file offset. -- Duy