On 10/23/18 9:06 AM, Jan Tulak wrote: > On Tue, Oct 23, 2018 at 3:43 PM Eric Sandeen <sandeen@xxxxxxxxxxx> wrote: >> >> On 10/23/18 8:15 AM, Jan Tulak wrote: >>> On Tue, Oct 23, 2018 at 2:10 PM Eric Sandeen <sandeen@xxxxxxxxxxx> wrote: > >>>> It's always possible that I'm wrong and I'm missing something. :) >>>> What did gcc say for the cases I pointed out? >>> >>> For the txt[256] -> 300: >>> >>> stobj.c: In function ‘stobjstrm_highlight’: >>> stobj.c:227:55: warning: ‘%s’ directive output may be truncated >>> writing up to 255 bytes into a region of size between 230 and 231 [-Wformat-truncation=] >>> snprintf(txt, sizeof(txt), "interrupted: %s, cmdarg: %s", >>> ^~ >>> stobj.c:227:2: note: ‘snprintf’ output between 26 and 282 bytes into a >>> destination of size 256 >>> snprintf(txt, sizeof(txt), "interrupted: %s, cmdarg: %s", >>> ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ >>> (stobjstrm->st_interrupted == BOOL_TRUE) ? "yes" : "no", >>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ >>> stobjstrm->st_cmdarg); >>> ~~~~~~~~~~~~~~~~~~~~~ >>> >>> So, I took the 282, added something, and rounded it to a number that >>> felt good and wasn't much larger than that. >> >> Ok, I don't think that's one of the ones I commented on ;) > > You mentioned multiple cases, so I picked the first one I found. :-) > If you ask for the 1024 -> 5000 bytes buffer, then it is probably this one: > > invidx.c:253:37: warning: ‘%s’ directive output may be truncated > writing up to 4095 bytes into a r$ > gion of size 1020 [-Wformat-truncation=] > snprintf(cmd, sizeof(cmd), "cp %s %s", stobjfile, dst_stobjfile); > ^~ ~~~~~~~~~~~~~ > invidx.c:253:3: note: ‘snprintf’ output 5 or more bytes (assuming > 4100) into a destination of size > 1024 > snprintf(cmd, sizeof(cmd), "cp %s %s", stobjfile, dst_stobjfile); > ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > > I checked the MAXPATHLEN and we have: > > ./man/man8/xfsrestore.8:775:must be 1023 characters (MAXPATHLEN) or less. > And also /common/fs.h:27:#define FS_MAXPATHLEN_DEFAULT 1024 Except checking at build time I get: invidx.c: In function ‘invidx_commit’: invidx.c:125: note: #pragma message: MAXPATHLEN=4096 > > The exact declaration of MAXPATHLEN is missing in my grep (probably > some magic created during build time), so, I'm not sure how exactly it > gets the 4100, but in other places, we repeatedly do things like > calloc( 1, 2 * MAXPATHLEN ), and if the default and manpage values are > ok and the 2* propagates here, it makes it 4096+4=4100 without \0. Right, so again, using the macro here makes it more foolproof. Thanks, -Eric
