| Message ID | 1572849196-21775-1-git-send-email-anand.jain@oracle.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | btrfs-progs: global verbose and quiet option | expand |
On Mon, Nov 04, 2019 at 02:32:58PM +0800, Anand Jain wrote: > v1->v1.1: > . Fix typo in HELPINFO_INSERT_QUIET. > . Remove #include <stdbool.h> where its no more required. > (was needed when %bconf.verbose was declared as bool). > . Use pr_verbose(-1,..) instead of all conditions printf() > . Use pr_verbose(1,..) instead of pr_verbose(true,..) > > verbosity sample code as in v1.1 Please user integer numbers for patch revisions, no need to mark patches that haven't changed, the overall summary of changes can mention what changed where if needed. > pr_verbose() > ------------ > /* > * level -1: prints message unless bconf.verbose == 0; > * level 0: quiet Are we ever going to call the function with level == 0? > * level >0: prints message only if <= bconf.verbose > */ > void pr_verbose(int level, const char *fmt, ...) > { > va_list args; > > if (level == 0 || bconf.verbose == 0) > return; > > if (level > bconf.verbose) > return; > > va_start(args, fmt); > vfprintf(stdout, fmt, args); > va_end(args); > } > > 1. > There are certain sub-commands which does not have any verbose output > or quiet output. However if the global options were used with those > sub-commands then the command shall not report any usage error. Or > my question is should it error out.? For example: > (with the patch) btrfs --verbose device ready /dev/sdb > actually there isn't any verbose output but we won't error out. > Similarly, > (without the patch) btrfs send -vvvvv will not show usage error > as well. > So I believe this is fine. IMO. Yes this is fine. > 2. > There is slight difference in output when global options are used > as compared to the output using the same sequence of options at the > sub-command level. For example: > > btrfs send -v -q -v is-equal-to btrfs send > But same sequence in the global option > btrfs -v -q -v send is-not-equal-to btrfs send > but is-equal-to btrfs -v send or btrfs send -v. > (similarly applies to receive as well). > > which IMO is fair expectation as -v is ending last. Agreed, hopefully the wild combinations of -v and -q are not too common. The patchset looks good, though it needs the small fixups like the global header or the helper macros, the core of the changes is there. That should be good for 5.4. The first version merged should bring the support for global verbosity options, then we can gradually convert all fprintf/printf to the helpers and add new printfs with higher verbosity levels.
On 11/16/19 12:11 AM, David Sterba wrote: > On Mon, Nov 04, 2019 at 02:32:58PM +0800, Anand Jain wrote: >> v1->v1.1: >> . Fix typo in HELPINFO_INSERT_QUIET. >> . Remove #include <stdbool.h> where its no more required. >> (was needed when %bconf.verbose was declared as bool). >> . Use pr_verbose(-1,..) instead of all conditions printf() >> . Use pr_verbose(1,..) instead of pr_verbose(true,..) >> >> verbosity sample code as in v1.1 > > Please user integer numbers for patch revisions, no need to mark patches > that haven't changed, the overall summary of changes can mention what > changed where if needed. Ok got it. (I kind of didn't want to integer++ unless it fixes review comments). >> pr_verbose() >> ------------ >> /* >> * level -1: prints message unless bconf.verbose == 0; >> * level 0: quiet > > Are we ever going to call the function with level == 0? Yes. For example in the patch [PATCH v1.1 13/18] btrfs-progs: restore: use global verbose option -------------- @@ -375,8 +374,7 @@ static int copy_one_extent(struct btrfs_root *root, int fd, if (compress == BTRFS_COMPRESS_NONE) bytenr += offset; - if (verbose && offset) - printf("offset is %Lu\n", offset); + pr_verbose(offset ? 1 : 0, "offset is %Lu\n", offset); -------------- >> * level >0: prints message only if <= bconf.verbose >> */ >> void pr_verbose(int level, const char *fmt, ...) >> { >> va_list args; >> >> if (level == 0 || bconf.verbose == 0) >> return; >> >> if (level > bconf.verbose) >> return; >> >> va_start(args, fmt); >> vfprintf(stdout, fmt, args); >> va_end(args); >> } >> >> 1. >> There are certain sub-commands which does not have any verbose output >> or quiet output. However if the global options were used with those >> sub-commands then the command shall not report any usage error. Or >> my question is should it error out.? For example: >> (with the patch) btrfs --verbose device ready /dev/sdb >> actually there isn't any verbose output but we won't error out. >> Similarly, >> (without the patch) btrfs send -vvvvv will not show usage error >> as well. >> So I believe this is fine. IMO. > > Yes this is fine. > >> 2. >> There is slight difference in output when global options are used >> as compared to the output using the same sequence of options at the >> sub-command level. For example: >> >> btrfs send -v -q -v is-equal-to btrfs send >> But same sequence in the global option >> btrfs -v -q -v send is-not-equal-to btrfs send >> but is-equal-to btrfs -v send or btrfs send -v. >> (similarly applies to receive as well). >> >> which IMO is fair expectation as -v is ending last. > > Agreed, hopefully the wild combinations of -v and -q are not too common. > > The patchset looks good, though it needs the small fixups like the > global header or the helper macros, the core of the changes is there. > That should be good for 5.4. > > The first version merged should bring the support for global verbosity > options, then we can gradually convert all fprintf/printf to the helpers > and add new printfs with higher verbosity levels. > Yes. Will do. Thanks, Anand
v1->v1.1: . Fix typo in HELPINFO_INSERT_QUIET. . Remove #include <stdbool.h> where its no more required. (was needed when %bconf.verbose was declared as bool). . Use pr_verbose(-1,..) instead of all conditions printf() . Use pr_verbose(1,..) instead of pr_verbose(true,..) verbosity sample code as in v1.1 global init ----------- bconf.verbose = -1; //-1:default, 0:quiet, >1:verbose at global options ----------------- case 'v': bconf.verbose < 0 ? bconf.verbose = 1 : bconf.verbose++; break; case 'q': bconf.verbose = 0; break; sub-command init ---------------- For send/receive only (special cases, default verbosity is 1): if (bconf.verbose < 0) bconf.verbose = 1; else if (bconf.verbose > 0) bconf.verbose++; Non-send/receive: default verbosity is 0 (ref: cmds/rescue.c) if (bconf.verbose < 0) bconf.verbose = 0; at sub-command options ---------------------- case 'v': bconf.verbose++; break; case 'q': bconf.verbose = 0; break; pr_verbose() ------------ /* * level -1: prints message unless bconf.verbose == 0; * level 0: quiet * level >0: prints message only if <= bconf.verbose */ void pr_verbose(int level, const char *fmt, ...) { va_list args; if (level == 0 || bconf.verbose == 0) return; if (level > bconf.verbose) return; va_start(args, fmt); vfprintf(stdout, fmt, args); va_end(args); } RFC->v1: . Adds --quiet option to the global btrfs(8) command. . Used global struct bconf. . Refactored pr_verbose(), accepts level (int) as argument, so now the sub-command can specify the verbose level at which the particular verbose messages has to be printed. . Added global help defines. Kindly note the following:- 1. There are certain sub-commands which does not have any verbose output or quiet output. However if the global options were used with those sub-commands then the command shall not report any usage error. Or my question is should it error out.? For example: (with the patch) btrfs --verbose device ready /dev/sdb actually there isn't any verbose output but we won't error out. Similarly, (without the patch) btrfs send -vvvvv will not show usage error as well. So I believe this is fine. IMO. 2. There is slight difference in output when global options are used as compared to the output using the same sequence of options at the sub-command level. For example: btrfs send -v -q -v is-equal-to btrfs send But same sequence in the global option btrfs -v -q -v send is-not-equal-to btrfs send but is-equal-to btrfs -v send or btrfs send -v. (similarly applies to receive as well). which IMO is fair expectation as -v is ending last. RFC: This patch set brings --verbose option to the top level btrfs command, such as 'btrfs --verbose'. With this we don't have to add or remember verbose option at the sub-commands level. As there are already verbose options to 11 sub-commands as listed below [1][2]. So the top level --verbose option here takes care to transpire verbose request from the top level to the sub-command level for 9 (not 11) sub-commands as in [1] as of now. This patch is RFC still for the following two reasons (comments appreciated). 1. The sub-commands as in [2] uses multi-level compile time verbose option, such as %g_verbose = 0 (quite), %g_verbose = 1 (default), %g_verbose > 1 (real-verbose). And verbose at default is also part the .out files in fstests. So it needs further discussions on how to handle the multi- level verbose option using the global verbose option, and so sub- commands in [2] are untouched. 2. These patch has been unit-tested individually. These patches does not alter the verbose output. But it fixes the indentation in the command's help output, which may be used in fstests and btrfs-progs/tests and their verification is pending still, which I am planning to do it before v1. [1] btrfs subvolume delete --help -v|--verbose verbose output of operations btrfs filesystem defragment --help -v be verbose btrfs balance start --help -v|--verbose be verbose btrfs balance status --help -v|--verbose be verbose btrfs rescue chunk-recover --help -v Verbose mode btrfs rescue super-recover --help -v Verbose mode btrfs restore --help -v|--verbose verbose btrfs inspect-internal inode-resolve --help -v verbose mode btrfs inspect-internal logical-resolve --help -v verbose mode [2] btrfs send --help -v|--verbose enable verbose output to stderr, each occurrence of btrfs receive --help -v increase verbosity about performed action Anand Jain (18): btrfs-progs: receive: fix option quiet btrfs-progs: balance status: fix usage show long verbose btrfs-progs: balance start: fix usage add long verbose btrfs-progs: add global verbose and quiet options and helper functions btrfs-progs: send: use global verbose and quiet options btrfs-progs: receive: use global verbose and quiet options btrfs-progs: subvolume delete: use global verbose option btrfs-progs: filesystem defragment: use global verbose option btrfs-progs: balance start: use global verbose option btrfs-progs: balance status: use global verbose option btrfs-progs: rescue chunk-recover: use global verbose option btrfs-progs: rescue super-recover: use global verbose option btrfs-progs: restore: use global verbose option btrfs-progs: inspect-internal inode-resolve: use global verbose btrfs-progs: inspect-internal logical-resolve: use global verbose option btrfs-progs: refactor btrfs_scan_devices() to accept verbose argument btrfs-progs: device scan: add verbose option btrfs-progs: device scan: add quiet option btrfs.c | 20 +++++++++++-- cmds/balance.c | 26 +++++++++++------ cmds/device.c | 7 +++-- cmds/filesystem.c | 18 ++++++------ cmds/inspect.c | 45 +++++++++++++++--------------- cmds/receive.c | 79 +++++++++++++++++++++++++++++----------------------- cmds/rescue.c | 22 +++++++++++---- cmds/restore.c | 57 ++++++++++++++++++------------------- cmds/send.c | 33 ++++++++++++++-------- cmds/subvolume.c | 32 +++++++++++---------- common/device-scan.c | 3 +- common/device-scan.h | 2 +- common/help.h | 11 ++++++++ common/messages.c | 17 +++++++++++ common/messages.h | 3 ++ common/utils.c | 3 +- common/utils.h | 3 ++ disk-io.c | 2 +- 18 files changed, 239 insertions(+), 144 deletions(-)