checkpatch: add fix option for NETWORKING_BLOCK_COMMENT_STYLE
PATCH 1/1 checkpatch: add fix option for NETWORKING_BLOCK_COMMENT_STYLE
Currently, checkpatch warns us if we use an empty '/*' line for comment and contents of the comment are in next line.
E.g., running checkpatch on commit 0d52497ac8ee ("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:
WARNING: networking block comments don't use an empty /* line, use /* Comment...
+ /*
+ * If the TXQ is active, then set the timer, if not,
Provide a fix by appending the current line contents to previous line and deleting the current line
--- scripts/checkpatch.pl | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)
scripts/checkpatch.pl +8 -2
@@ -3629,8 +3629,14 @@ sub process { $prevrawline =~ /^\+[ \t]*\/\*[ \t]*$/ && $rawline =~ /^\+[ \t]*\*/ && $realline > 3) { # Do not warn about the initial copyright comment block after SPDX-License-Identifier - WARN("NETWORKING_BLOCK_COMMENT_STYLE", - "networking block comments don't use an empty /* line, use /* Comment...\n" . $hereprev); + if (WARN("NETWORKING_BLOCK_COMMENT_STYLE", + "networking block comments don't use an empty /* line, use /* Comment...\n" . $hereprev) &&\ + $fix) { + if ($rawline =~ /^\+[ \t]*\*\s*(.*)/) { + fix_delete_line($fixlinenr, $rawline); + $fixed[$fixlinenr - 1] .= " $1"; + } + } } # Block comments use * on subsequent lines
Signature · 6 lines
2.17.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 30 lines · Aditya Srivastava
Currently, checkpatch warns us if we use an empty '/*' line for comment and contents of the comment are in next line.
E.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Provide a fix by appending the current line contents to previous line and deleting the current line
Signed-off-by Aditya Srivastava <[email protected]>--- scripts/checkpatch.pl | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)scripts/checkpatch.pl +1
-2@@ -3629,8 +3629,14 @@ sub process { $prevrawline =~ /^\+[ \t]*\/\*[ \t]*$/ && $rawline =~ /^\+[ \t]*\*/ && $realline > 3) { # Do not warn about the initial copyright comment block after SPDX-License-Identifier - WARN("NETWORKING_BLOCK_COMMENT_STYLE", - "networking block comments don't use an empty /* line, use /* Comment...\n" . $hereprev); + if (WARN("NETWORKING_BLOCK_COMMENT_STYLE",
scripts/checkpatch.pl sub process { · line 3642
Quoted 10 lines
scripts/checkpatch.pl +7
-0+ "networking block comments don't use an empty /* line, use /* Comment...\n" . $hereprev) &&\ + $fix) { + if ($rawline =~ /^\+[ \t]*\*\s*(.*)/) { + fix_delete_line($fixlinenr, $rawline); + $fixed[$fixlinenr - 1] .= " $1"; + } + } } # Block comments use * on subsequent lines
I have generated the warnings occurring for this class of warnings (over 4.13..5.8):
https://github.com/AdityaSrivast/kernel-tasks/blob/master/random/networking_block_comments/warnings.txt
My initial tests at these warnings found the fix to be working fine.
Thanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH v2 1/1 checkpatch: add fix option for NETWORKING_BLOCK_COMMENT_STYLE
Currently, checkpatch warns us for files in 'net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next line
E.g., running checkpatch on commit 0d52497ac8ee ("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:
WARNING: networking block comments don't use an empty /* line, use /* Comment...
+ /*
+ * If the TXQ is active, then set the timer, if not,
Provide a fix by appending the current line contents to previous line and removing the current line
---
Changes in v2: modify commit message: add information about networking files in commit message
scripts/checkpatch.pl | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)
scripts/checkpatch.pl +8 -2
@@ -3629,8 +3629,14 @@ sub process { $prevrawline =~ /^\+[ \t]*\/\*[ \t]*$/ && $rawline =~ /^\+[ \t]*\*/ && $realline > 3) { # Do not warn about the initial copyright comment block after SPDX-License-Identifier - WARN("NETWORKING_BLOCK_COMMENT_STYLE", - "networking block comments don't use an empty /* line, use /* Comment...\n" . $hereprev); + if (WARN("NETWORKING_BLOCK_COMMENT_STYLE", + "networking block comments don't use an empty /* line, use /* Comment...\n" . $hereprev) &&\ + $fix) { + if ($rawline =~ /^\+[ \t]*\*\s*(.*)/) { + fix_delete_line($fixlinenr, $rawline); + $fixed[$fixlinenr - 1] .= " $1"; + } + } } # Block comments use * on subsequent lines
Signature · 6 lines
2.17.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
Lukas
Quoted 34 lines
Signed-off-by Aditya Srivastava <[email protected]>---
Changes in v2: modify commit message: add information about networking files in commit messagescripts/checkpatch.pl | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)scripts/checkpatch.pl +0
-0@@ -3629,8 +3629,14 @@ sub process { $prevrawline =~ /^\+[ \t]*\/\*[ \t]*$/ && $rawline =~ /^\+[ \t]*\*/ && $realline > 3) { # Do not warn about the initialcopyright comment block after SPDX-License-Identifier
scripts/checkpatch.pl +0
-2- WARN("NETWORKING_BLOCK_COMMENT_STYLE", - "networking block comments don't use an empty/* line, use /* Comment...\n" . $hereprev);
scripts/checkpatch.pl +2
-0+ if (WARN("NETWORKING_BLOCK_COMMENT_STYLE", + "networking block comments don't use anempty /* line, use /* Comment...\n" . $hereprev) &&\
scripts/checkpatch.pl +3
-0+ $fix) { + if ($rawline =~ /^\+[ \t]*\*\s*(.*)/) { + fix_delete_line($fixlinenr,$rawline);
scripts/checkpatch.pl +3
-1+ $fixed[$fixlinenr - 1] .= " $1"; + } + } } # Block comments use * on subsequent lines --
Quoted 3 lines
2.17.1
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,
Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
In drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)
In net/:
Wrong style: 5090 lines
Correct style: 30485 lines
There is also a documentation regarding this different comment format, and can be found here:
https://www.kernel.org/doc/html/latest/process/coding-style.html#commenting
Thanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 35 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
In
drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)In
net/:
Wrong style: 5090 lines
Correct style: 30485 lines
Quoted 7 lines
There is also a documentation regarding this different comment format, and can be found here:
https://www.kernel.org/doc/html/latest/process/coding-style.html#commentingThanks
Aditya
Hi Lukas
This mail probably got missed. Please review this as well :)
Thanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 28 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
Quoted 8 lines
In
drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)In
net/:
Wrong style: 5090 lines
Correct style: 30485 lines
Can you find out where the wrong style is used?
Maybe the documentation is a bit outdated.
For example, some drivers and some subdirectories might have settled for another commenting style.
I guess you can submit the fix option, but it could be that the whole rule/feature is broken anyway... so I do not know if that fix is of any good...
I think it is better to actually understand, document and encode the current rule that applies.
Can you provide an evaluation where the different styles for commenting are aggregated in a good way?
E.g., consistently within a file with style XYZ; mixed style but 80-90% are of style ABC; consistent within a directory.
If you think some cases are in the wrong style for some specific files, simply send a patch correcting the commenting style and see.
Lukas
Quoted 6 lines
There is also a documentation regarding this different comment format, and can be found here:
https://www.kernel.org/doc/html/latest/process/coding-style.html#commentingThanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 52 lines · Lukas Bulwahn
Quoted 28 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
Quoted 8 lines
In
drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)In
net/:
Wrong style: 5090 lines
Correct style: 30485 linesCan you find out where the wrong style is used?
Maybe the documentation is a bit outdated.
For example, some drivers and some subdirectories might have settled for another commenting style.
I guess you can submit the fix option, but it could be that the whole rule/feature is broken anyway... so I do not know if that fix is of any good...
I think it is better to actually understand, document and encode the current rule that applies.
Quoted 8 lines
Can you provide an evaluation where the different styles for commenting are aggregated in a good way?
E.g., consistently within a file with style XYZ; mixed style but 80-90% are of style ABC; consistent within a directory.If you think some cases are in the wrong style for some specific files, simply send a patch correcting the commenting style and see.
Hi Lukas
I have generated the reports regarding the comments used in various files at net/ and drivers/net.
Directory wise reports:
File wise reports:
Some files in 'net/' follow the accepted syntax, such as:
'net/batman-adv => 985 0'.
However, some completely not, such as:
'net/ax25 => 0 103'.
Some even follow mixed syntax such as:
'net/netfilter => 87 125'.
Similarly for drivers/net:drivers/net/ethernet/mellanox/mlxsw => 1071 0
completely follow accepted syntax.
drivers/net/wireless/ralink/rt2x00 => 98, 2082
follow unaccepted syntax largely.
whereas files such as drivers/net/wireless/ath/ath5k => 304, 486 follow mixed syntax.
It seems to me as the documentation might be outdated, as you had also suggested earlier, or maybe users are unaware of the documentation and
just use the syntax for conserving consistency of the code.
What do you think?
Thanks
Aditya
Quoted 2 lines
Lukas
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 67 lines · Aditya
Quoted 52 lines · Lukas Bulwahn
Quoted 28 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
Quoted 8 lines
In
drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)In
net/:
Wrong style: 5090 lines
Correct style: 30485 linesCan you find out where the wrong style is used?
Maybe the documentation is a bit outdated.
For example, some drivers and some subdirectories might have settled for another commenting style.
I guess you can submit the fix option, but it could be that the whole rule/feature is broken anyway... so I do not know if that fix is of any good...
I think it is better to actually understand, document and encode the current rule that applies.
Quoted 8 lines
Can you provide an evaluation where the different styles for commenting are aggregated in a good way?
E.g., consistently within a file with style XYZ; mixed style but 80-90% are of style ABC; consistent within a directory.If you think some cases are in the wrong style for some specific files, simply send a patch correcting the commenting style and see.
Hi Lukas
I have generated the reports regarding the comments used in various files atnet/anddrivers/net.
Directory wise reports:
Quoted 6 lines
Can you sort that by the overall number of comments per directory?
Quoted 24 lines
File wise reports:
Some files in '
net/' follow the accepted syntax, such as:
'net/batman-adv=> 985 0'.However, some completely not, such as:
'net/ax25 => 0 103'.Some even follow mixed syntax such as:
'net/netfilter => 87 125'.Similarly for
drivers/net:drivers/net/ethernet/mellanox/mlxsw=> 1071 0
completely follow accepted syntax.
drivers/net/wireless/ralink/rt2x00=> 98, 2082
follow unaccepted syntax largely.
Quoted 9 lines
whereas files such as
drivers/net/wireless/ath/ath5k=> 304, 486 follow mixed syntax.It seems to me as the documentation might be outdated, as you had also suggested earlier, or maybe users are unaware of the documentation and
just use the syntax for conserving consistency of the code.What do you think?
I think it largely depends on the history of the code and developers that take care...
We might just start to ask the larger areas to modify the rule for their directory.
Lukas
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 109 lines · Lukas Bulwahn
Quoted 67 lines · Aditya
Quoted 52 lines · Lukas Bulwahn
Quoted 28 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
Quoted 8 lines
In
drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)In
net/:
Wrong style: 5090 lines
Correct style: 30485 linesCan you find out where the wrong style is used?
Maybe the documentation is a bit outdated.
For example, some drivers and some subdirectories might have settled for another commenting style.
I guess you can submit the fix option, but it could be that the whole rule/feature is broken anyway... so I do not know if that fix is of any good...
I think it is better to actually understand, document and encode the current rule that applies.
Quoted 8 lines
Can you provide an evaluation where the different styles for commenting are aggregated in a good way?
E.g., consistently within a file with style XYZ; mixed style but 80-90% are of style ABC; consistent within a directory.If you think some cases are in the wrong style for some specific files, simply send a patch correcting the commenting style and see.
Hi Lukas
I have generated the reports regarding the comments used in various files atnet/anddrivers/net.
Directory wise reports:Quoted 6 lines
Can you sort that by the overall number of comments per directory?
Quoted 31 lines
File wise reports:
Some files in '
net/' follow the accepted syntax, such as:
'net/batman-adv=> 985 0'.However, some completely not, such as:
'net/ax25 => 0 103'.Some even follow mixed syntax such as:
'net/netfilter => 87 125'.Similarly for
drivers/net:drivers/net/ethernet/mellanox/mlxsw=> 1071 0
completely follow accepted syntax.
drivers/net/wireless/ralink/rt2x00=> 98, 2082
follow unaccepted syntax largely.whereas files such as
drivers/net/wireless/ath/ath5k=> 304, 486 follow mixed syntax.It seems to me as the documentation might be outdated, as you had also suggested earlier, or maybe users are unaware of the documentation and
just use the syntax for conserving consistency of the code.
Quoted 9 lines
Quoted 2 lines
What do you think?
I think it largely depends on the history of the code and developers that take care...
We might just start to ask the larger areas to modify the rule for their directory.
I suggest we start with looking at the (let us say 10) "largest" directories with respect to number of comment blocks and then decide on a refinement of the rule for those subdirectories:
A. Directory follows NETWORKING COMMENT STYLE; checkpatch shall warn to follow NETWORKING COMMENT STYLE.
B. Directory follows GENERAL KERNEL COMMENT STYLE; checkpatch shall warn to follow GENERAL KERNEL COMMENT STYLE.
C. Directory is too mixed; checkpatch shall not warn on any comment style.
We can then start the discussion with the appropriate maintainer (and check what else this maintainer might maintain and if that is consistent) if they agree or disagree on that.
Lukas
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 130 lines · Lukas Bulwahn
Quoted 109 lines · Lukas Bulwahn
Quoted 67 lines · Aditya
Quoted 52 lines · Lukas Bulwahn
Quoted 28 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
Quoted 8 lines
In
drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)In
net/:
Wrong style: 5090 lines
Correct style: 30485 linesCan you find out where the wrong style is used?
Maybe the documentation is a bit outdated.
For example, some drivers and some subdirectories might have settled for another commenting style.
I guess you can submit the fix option, but it could be that the whole rule/feature is broken anyway... so I do not know if that fix is of any good...
I think it is better to actually understand, document and encode the current rule that applies.
Quoted 8 lines
Can you provide an evaluation where the different styles for commenting are aggregated in a good way?
E.g., consistently within a file with style XYZ; mixed style but 80-90% are of style ABC; consistent within a directory.If you think some cases are in the wrong style for some specific files, simply send a patch correcting the commenting style and see.
Hi Lukas
I have generated the reports regarding the comments used in various files atnet/anddrivers/net.
Directory wise reports:Quoted 6 lines
Can you sort that by the overall number of comments per directory?
Quoted 31 lines
File wise reports:
Some files in '
net/' follow the accepted syntax, such as:
'net/batman-adv=> 985 0'.However, some completely not, such as:
'net/ax25 => 0 103'.Some even follow mixed syntax such as:
'net/netfilter => 87 125'.Similarly for
drivers/net:drivers/net/ethernet/mellanox/mlxsw=> 1071 0
completely follow accepted syntax.
drivers/net/wireless/ralink/rt2x00=> 98, 2082
follow unaccepted syntax largely.whereas files such as
drivers/net/wireless/ath/ath5k=> 304, 486 follow mixed syntax.It seems to me as the documentation might be outdated, as you had also suggested earlier, or maybe users are unaware of the documentation and
just use the syntax for conserving consistency of the code.Quoted 9 lines
Quoted 2 lines
What do you think?
I think it largely depends on the history of the code and developers that take care...
We might just start to ask the larger areas to modify the rule for their directory.
I suggest we start with looking at the (let us say 10) "largest" directories with respect to number of comment blocks and then decide on a refinement of the rule for those subdirectories:
A. Directory follows NETWORKING COMMENT STYLE; checkpatch shall warn to follow NETWORKING COMMENT STYLE.
B. Directory follows GENERAL KERNEL COMMENT STYLE; checkpatch shall warn to follow GENERAL KERNEL COMMENT STYLE.
C. Directory is too mixed; checkpatch shall not warn on any comment style.
Quoted 4 lines
We can then start the discussion with the appropriate maintainer (and check what else this maintainer might maintain and if that is consistent) if they agree or disagree on that.
Hi Lukas
Here is the list I have generated with the directories sorted on the basis of comment count and labelled the top 10 directories as following NETWORKING, GENERIC or MIXED comment style(s):
At "drivers/net":
https://github.com/AdityaSrivast/kernel-tasks/blob/master/random/networking_block_comments/drivers_net/cumulative_sorted.txt
What do you think?
Thanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 142 lines · Aditya
Quoted 130 lines · Lukas Bulwahn
Quoted 109 lines · Lukas Bulwahn
Quoted 67 lines · Aditya
Quoted 52 lines · Lukas Bulwahn
Quoted 28 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
Quoted 8 lines
In
drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)In
net/:
Wrong style: 5090 lines
Correct style: 30485 linesCan you find out where the wrong style is used?
Maybe the documentation is a bit outdated.
For example, some drivers and some subdirectories might have settled for another commenting style.
I guess you can submit the fix option, but it could be that the whole rule/feature is broken anyway... so I do not know if that fix is of any good...
I think it is better to actually understand, document and encode the current rule that applies.
Quoted 8 lines
Can you provide an evaluation where the different styles for commenting are aggregated in a good way?
E.g., consistently within a file with style XYZ; mixed style but 80-90% are of style ABC; consistent within a directory.If you think some cases are in the wrong style for some specific files, simply send a patch correcting the commenting style and see.
Hi Lukas
I have generated the reports regarding the comments used in various files atnet/anddrivers/net.
Directory wise reports:Quoted 6 lines
Can you sort that by the overall number of comments per directory?
Quoted 31 lines
File wise reports:
Some files in '
net/' follow the accepted syntax, such as:
'net/batman-adv=> 985 0'.However, some completely not, such as:
'net/ax25 => 0 103'.Some even follow mixed syntax such as:
'net/netfilter => 87 125'.Similarly for
drivers/net:drivers/net/ethernet/mellanox/mlxsw=> 1071 0
completely follow accepted syntax.
drivers/net/wireless/ralink/rt2x00=> 98, 2082
follow unaccepted syntax largely.whereas files such as
drivers/net/wireless/ath/ath5k=> 304, 486 follow mixed syntax.It seems to me as the documentation might be outdated, as you had also suggested earlier, or maybe users are unaware of the documentation and
just use the syntax for conserving consistency of the code.Quoted 9 lines
Quoted 2 lines
What do you think?
I think it largely depends on the history of the code and developers that take care...
We might just start to ask the larger areas to modify the rule for their directory.
I suggest we start with looking at the (let us say 10) "largest" directories with respect to number of comment blocks and then decide on a refinement of the rule for those subdirectories:
A. Directory follows NETWORKING COMMENT STYLE; checkpatch shall warn to follow NETWORKING COMMENT STYLE.
B. Directory follows GENERAL KERNEL COMMENT STYLE; checkpatch shall warn to follow GENERAL KERNEL COMMENT STYLE.
C. Directory is too mixed; checkpatch shall not warn on any comment style.Quoted 4 lines
We can then start the discussion with the appropriate maintainer (and check what else this maintainer might maintain and if that is consistent) if they agree or disagree on that.
Hi Lukas
Here is the list I have generated with the directories sorted on the basis of comment count and labelled the top 10 directories as following NETWORKING, GENERIC or MIXED comment style(s):
Quoted 8 lines
At "
drivers/net":
https://github.com/AdityaSrivast/kernel-tasks/blob/master/random/networking_block_comments/drivers_net/cumulative_sorted.txtWhat do you think?
Obviously, this existing rule needs to be refined.
Do you have a proposal? Probably, you would like to have each maintainer state which comment style is preferred or so.
How could this be technically done?
Lukas
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 152 lines · Lukas Bulwahn
Quoted 142 lines · Aditya
Quoted 130 lines · Lukas Bulwahn
Quoted 109 lines · Lukas Bulwahn
Quoted 67 lines · Aditya
Quoted 52 lines · Lukas Bulwahn
Quoted 28 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
Quoted 8 lines
In
drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)In
net/:
Wrong style: 5090 lines
Correct style: 30485 linesCan you find out where the wrong style is used?
Maybe the documentation is a bit outdated.
For example, some drivers and some subdirectories might have settled for another commenting style.
I guess you can submit the fix option, but it could be that the whole rule/feature is broken anyway... so I do not know if that fix is of any good...
I think it is better to actually understand, document and encode the current rule that applies.
Quoted 8 lines
Can you provide an evaluation where the different styles for commenting are aggregated in a good way?
E.g., consistently within a file with style XYZ; mixed style but 80-90% are of style ABC; consistent within a directory.If you think some cases are in the wrong style for some specific files, simply send a patch correcting the commenting style and see.
Hi Lukas
I have generated the reports regarding the comments used in various files atnet/anddrivers/net.
Directory wise reports:Quoted 6 lines
Can you sort that by the overall number of comments per directory?
Quoted 31 lines
File wise reports:
Some files in '
net/' follow the accepted syntax, such as:
'net/batman-adv=> 985 0'.However, some completely not, such as:
'net/ax25 => 0 103'.Some even follow mixed syntax such as:
'net/netfilter => 87 125'.Similarly for
drivers/net:drivers/net/ethernet/mellanox/mlxsw=> 1071 0
completely follow accepted syntax.
drivers/net/wireless/ralink/rt2x00=> 98, 2082
follow unaccepted syntax largely.whereas files such as
drivers/net/wireless/ath/ath5k=> 304, 486 follow mixed syntax.It seems to me as the documentation might be outdated, as you had also suggested earlier, or maybe users are unaware of the documentation and
just use the syntax for conserving consistency of the code.Quoted 9 lines
Quoted 2 lines
What do you think?
I think it largely depends on the history of the code and developers that take care...
We might just start to ask the larger areas to modify the rule for their directory.
I suggest we start with looking at the (let us say 10) "largest" directories with respect to number of comment blocks and then decide on a refinement of the rule for those subdirectories:
A. Directory follows NETWORKING COMMENT STYLE; checkpatch shall warn to follow NETWORKING COMMENT STYLE.
B. Directory follows GENERAL KERNEL COMMENT STYLE; checkpatch shall warn to follow GENERAL KERNEL COMMENT STYLE.
C. Directory is too mixed; checkpatch shall not warn on any comment style.Quoted 4 lines
We can then start the discussion with the appropriate maintainer (and check what else this maintainer might maintain and if that is consistent) if they agree or disagree on that.
Hi Lukas
Here is the list I have generated with the directories sorted on the basis of comment count and labelled the top 10 directories as following NETWORKING, GENERIC or MIXED comment style(s):
Quoted 7 lines
Obviously, this existing rule needs to be refined.
Do you have a proposal? Probably, you would like to have each maintainer state which comment style is preferred or so.
How could this be technically done?
We can probably Cc all the corresponding maintainers checking from the MAINTAINERS file.
Since we are considering only the top 10 directories, this probably shouldn't be much time consuming.
Or we can probably mail any central maintainer(s) for all net and drivers/net directories (for eg, someone who decided the comment rule in the beginning for both the directories) and discuss.
What do you think?
Thanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 162 lines · Aditya
Quoted 152 lines · Lukas Bulwahn
Quoted 142 lines · Aditya
Quoted 130 lines · Lukas Bulwahn
Quoted 109 lines · Lukas Bulwahn
Quoted 67 lines · Aditya
Quoted 52 lines · Lukas Bulwahn
Quoted 28 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
Quoted 8 lines
In
drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)In
net/:
Wrong style: 5090 lines
Correct style: 30485 linesCan you find out where the wrong style is used?
Maybe the documentation is a bit outdated.
For example, some drivers and some subdirectories might have settled for another commenting style.
I guess you can submit the fix option, but it could be that the whole rule/feature is broken anyway... so I do not know if that fix is of any good...
I think it is better to actually understand, document and encode the current rule that applies.
Quoted 8 lines
Can you provide an evaluation where the different styles for commenting are aggregated in a good way?
E.g., consistently within a file with style XYZ; mixed style but 80-90% are of style ABC; consistent within a directory.If you think some cases are in the wrong style for some specific files, simply send a patch correcting the commenting style and see.
Hi Lukas
I have generated the reports regarding the comments used in various files atnet/anddrivers/net.
Directory wise reports:Quoted 6 lines
Can you sort that by the overall number of comments per directory?
Quoted 31 lines
File wise reports:
Some files in '
net/' follow the accepted syntax, such as:
'net/batman-adv=> 985 0'.However, some completely not, such as:
'net/ax25 => 0 103'.Some even follow mixed syntax such as:
'net/netfilter => 87 125'.Similarly for
drivers/net:drivers/net/ethernet/mellanox/mlxsw=> 1071 0
completely follow accepted syntax.
drivers/net/wireless/ralink/rt2x00=> 98, 2082
follow unaccepted syntax largely.whereas files such as
drivers/net/wireless/ath/ath5k=> 304, 486 follow mixed syntax.It seems to me as the documentation might be outdated, as you had also suggested earlier, or maybe users are unaware of the documentation and
just use the syntax for conserving consistency of the code.Quoted 9 lines
Quoted 2 lines
What do you think?
I think it largely depends on the history of the code and developers that take care...
We might just start to ask the larger areas to modify the rule for their directory.
I suggest we start with looking at the (let us say 10) "largest" directories with respect to number of comment blocks and then decide on a refinement of the rule for those subdirectories:
A. Directory follows NETWORKING COMMENT STYLE; checkpatch shall warn to follow NETWORKING COMMENT STYLE.
B. Directory follows GENERAL KERNEL COMMENT STYLE; checkpatch shall warn to follow GENERAL KERNEL COMMENT STYLE.
C. Directory is too mixed; checkpatch shall not warn on any comment style.Quoted 4 lines
We can then start the discussion with the appropriate maintainer (and check what else this maintainer might maintain and if that is consistent) if they agree or disagree on that.
Hi Lukas
Here is the list I have generated with the directories sorted on the basis of comment count and labelled the top 10 directories as following NETWORKING, GENERIC or MIXED comment style(s):Quoted 7 lines
Obviously, this existing rule needs to be refined.
Do you have a proposal? Probably, you would like to have each maintainer state which comment style is preferred or so.
How could this be technically done?
Quoted 9 lines
We can probably Cc all the corresponding maintainers checking from the
MAINTAINERSfile.
Since we are considering only the top 10 directories, this probably shouldn't be much time consuming.Or we can probably mail any central maintainer(s) for all net and
drivers/netdirectories (for eg, someone who decided the comment rule in the beginning for both the directories) and discuss.
It makes sense to identify the names of those people to start with; I would suggest NOT to contact them until we have a first good patch set to provide some technical solution.
I was asking more about how a maintainer can state their preference in the repository and how will checkpatch then use this information?
Did you think about a technical solution for that?
Lukas
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 173 lines · Lukas Bulwahn
Quoted 162 lines · Aditya
Quoted 152 lines · Lukas Bulwahn
Quoted 142 lines · Aditya
Quoted 130 lines · Lukas Bulwahn
Quoted 109 lines · Lukas Bulwahn
Quoted 67 lines · Aditya
Quoted 52 lines · Lukas Bulwahn
Quoted 28 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 12 lines · Aditya Srivastava
Currently, checkpatch warns us for files in '
net/' and 'drivers/net', if we use an empty '/*' line for comment and contents of comments are in next lineE.g., running checkpatch on commit
0d52497ac8ee("iwlwifi: pcie: remove the refs / unrefs from the transport") reports this warning:WARNING: networking block comments don't use an empty /* line, use /* Comment...
(quoted diff) +2
-0+ /* + * If the TXQ is active, then set the timer, if not,Quoted 10 lines
Quoted 3 lines
Provide a fix by appending the current line contents to previous line and removing the current line
Patch generally looks good.
Can you check how many comments in net actually follow that style and how many follow another style?
Quoted 8 lines
In
drivers/net:
Wrong style: 14695 lines
Correct style: 147961 lines (ie around 10 times)In
net/:
Wrong style: 5090 lines
Correct style: 30485 linesCan you find out where the wrong style is used?
Maybe the documentation is a bit outdated.
For example, some drivers and some subdirectories might have settled for another commenting style.
I guess you can submit the fix option, but it could be that the whole rule/feature is broken anyway... so I do not know if that fix is of any good...
I think it is better to actually understand, document and encode the current rule that applies.
Quoted 8 lines
Can you provide an evaluation where the different styles for commenting are aggregated in a good way?
E.g., consistently within a file with style XYZ; mixed style but 80-90% are of style ABC; consistent within a directory.If you think some cases are in the wrong style for some specific files, simply send a patch correcting the commenting style and see.
Hi Lukas
I have generated the reports regarding the comments used in various files atnet/anddrivers/net.
Directory wise reports:Quoted 6 lines
Can you sort that by the overall number of comments per directory?
Quoted 31 lines
File wise reports:
Some files in '
net/' follow the accepted syntax, such as:
'net/batman-adv=> 985 0'.However, some completely not, such as:
'net/ax25 => 0 103'.Some even follow mixed syntax such as:
'net/netfilter => 87 125'.Similarly for
drivers/net:drivers/net/ethernet/mellanox/mlxsw=> 1071 0
completely follow accepted syntax.
drivers/net/wireless/ralink/rt2x00=> 98, 2082
follow unaccepted syntax largely.whereas files such as
drivers/net/wireless/ath/ath5k=> 304, 486 follow mixed syntax.It seems to me as the documentation might be outdated, as you had also suggested earlier, or maybe users are unaware of the documentation and
just use the syntax for conserving consistency of the code.Quoted 9 lines
Quoted 2 lines
What do you think?
I think it largely depends on the history of the code and developers that take care...
We might just start to ask the larger areas to modify the rule for their directory.
I suggest we start with looking at the (let us say 10) "largest" directories with respect to number of comment blocks and then decide on a refinement of the rule for those subdirectories:
A. Directory follows NETWORKING COMMENT STYLE; checkpatch shall warn to follow NETWORKING COMMENT STYLE.
B. Directory follows GENERAL KERNEL COMMENT STYLE; checkpatch shall warn to follow GENERAL KERNEL COMMENT STYLE.
C. Directory is too mixed; checkpatch shall not warn on any comment style.Quoted 4 lines
We can then start the discussion with the appropriate maintainer (and check what else this maintainer might maintain and if that is consistent) if they agree or disagree on that.
Hi Lukas
Here is the list I have generated with the directories sorted on the basis of comment count and labelled the top 10 directories as following NETWORKING, GENERIC or MIXED comment style(s):Quoted 7 lines
Obviously, this existing rule needs to be refined.
Do you have a proposal? Probably, you would like to have each maintainer state which comment style is preferred or so.
How could this be technically done?
Quoted 9 lines
We can probably Cc all the corresponding maintainers checking from the
MAINTAINERSfile.
Since we are considering only the top 10 directories, this probably shouldn't be much time consuming.Or we can probably mail any central maintainer(s) for all net and
drivers/netdirectories (for eg, someone who decided the comment rule in the beginning for both the directories) and discuss.
Quoted 9 lines
It makes sense to identify the names of those people to start with; I would suggest NOT to contact them until we have a first good patch set to provide some technical solution.
I was asking more about how a maintainer can state their preference in the repository and how will checkpatch then use this information?
Did you think about a technical solution for that?
Yes, you are correct. We can probably create a file where we can store the directories(names) which are inside net or drivers/net, but do not follow NETWORKING_BLOCK_COMMENT style (something like networking_block_comment_style.ignore file). Maintainers can add their directories inside this file.
We can then ignore these files from this class of warning.
What do you think?
Thanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH 1/1 checkpatch: ignore files not following NETWORKING_BLOCK_COMMENT_STYLE
In reply to Lukas BulwahnCurrently checkpatch.pl gives warning for NETWORKING_BLOCK_COMMENT_STYLE for files in net/ and drivers/net which do not follow the networking comment style.
But some of these files seem to follow the generic comment style instead of networking style and some rather mixed style of comment.
For e.g., drivers/net/wireless/ralink/rt2x00 largely follows generic kernel comment style in spite of being inside drivers/net.
Provide an ignore file(".networking_block_comment_styles.ignore"), where users can add the files they want to ignore this warning.
--- .gitignore | 1 + .networking_block_comment_styles.ignore | 1 + scripts/checkpatch.pl | 25 ++++++++++++++++++++++++- 3 files changed, 26 insertions(+), 1 deletion(-) create mode 100644 .networking_block_comment_styles.ignore
.gitignore +1 -0
@@ -94,6 +94,7 @@ modules.order !.gitattributes !.gitignore !.mailmap +!.networking_block_comment_styles.ignore # # Generated include files
.networking_block_comment_styles.ignore new file +1 -0
@@ -0,0 +1 @@ +drivers/net/wireless/ralink/rt2x00
scripts/checkpatch.pl +24 -1
@@ -832,6 +832,28 @@ sub read_words { return 0; } +my @files_ignoring_networking_comment; +my $ignore_file = which_conf(".networking_block_comment_styles.ignore"); +if(-f $ignore_file) { + open(my $ignore, '<', "$ignore_file") + or warn "$P: Can't find a readable .networking_block_comment_styles.ignore file $!\n"; + while(<$ignore>) { + my $line = "$_"; + $line = trim($line); + next if ($line =~ /^\s*$/); + push(@files_ignoring_networking_comment, $line); + } + close($ignore); +} + +sub ignore_networking_comment_style { + my ($realfile) = "@_"; + foreach my $file (@files_ignoring_networking_comment) { + return 1 if ($realfile =~ m@^$file@); + } + return 0; +} + my $const_structs; if (show_type("CONST_STRUCT")) { read_words(\$const_structs, $conststructsfile)
@@ -3701,7 +3723,8 @@ sub process { if ($realfile =~ m@^(drivers/net/|net/)@ && $prevrawline =~ /^\+[ \t]*\/\*[ \t]*$/ && $rawline =~ /^\+[ \t]*\*/ && - $realline > 3) { # Do not warn about the initial copyright comment block after SPDX-License-Identifier + $realline > 3 && # Do not warn about the initial copyright comment block after SPDX-License-Identifier + !ignore_networking_comment_style($realfile)) { WARN("NETWORKING_BLOCK_COMMENT_STYLE", "networking block comments don't use an empty /* line, use /* Comment...\n" . $hereprev); }
Signature · 6 lines
2.17.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Re: [Linux-kernel-mentees] [PATCH] checkpatch: ignore files not following NETWORKING_BLOCK_COMMENT_STYLE
Quoted 75 lines · Aditya Srivastava
Currently checkpatch.pl gives warning for
NETWORKING_BLOCK_COMMENT_STYLEfor files innet/anddrivers/netwhich do not follow the networking comment style.
But some of these files seem to follow the generic comment style instead of networking style and some rather mixed style of comment.For e.g.,
drivers/net/wireless/ralink/rt2x00largely follows generic kernel comment style in spite of being insidedrivers/net.Provide an ignore
file(".networking_block_comment_styles.ignore"), where users can add the files they want to ignore this warning.Signed-off-by Aditya Srivastava <[email protected]>--- .gitignore | 1 + .networking_block_comment_styles.ignore | 1 + scripts/checkpatch.pl | 25 ++++++++++++++++++++++++- 3 files changed, 26 insertions(+), 1 deletion(-) create mode 100644 .networking_block_comment_styles.ignore.gitignore +1
-0@@ -94,6 +94,7 @@ modules.order !.gitattributes !.gitignore !.mailmap +!.networking_block_comment_styles.ignore # # Generated include files.networking_block_comment_styles.ignore new file +1
-0@@ -0,0 +1 @@ +drivers/net/wireless/ralink/rt2x00scripts/checkpatch.pl +22
-0@@ -832,6 +832,28 @@ sub read_words { return 0; } +my @files_ignoring_networking_comment; +my $ignore_file = which_conf(".networking_block_comment_styles.ignore"); +if(-f $ignore_file) { + open(my $ignore, '<', "$ignore_file") + or warn "$P: Can't find a readable .networking_block_comment_styles.ignore file $!\n"; + while(<$ignore>) { + my $line = "$_"; + $line = trim($line); + next if ($line =~ /^\s*$/); + push(@files_ignoring_networking_comment, $line); + } + close($ignore); +} + +sub ignore_networking_comment_style { + my ($realfile) = "@_"; + foreach my $file (@files_ignoring_networking_comment) { + return 1 if ($realfile =~ m@^$file@); + } + return 0; +} + my $const_structs; if (show_type("CONST_STRUCT")) { read_words(\$const_structs, $conststructsfile)
scripts/checkpatch.pl sub process { · line 3730
Quoted 9 lines
scripts/checkpatch.pl +2
-1@@ -3701,7 +3723,8 @@ sub process { if ($realfile =~ m@^(drivers/net/|net/)@ && $prevrawline =~ /^\+[ \t]*\/\*[ \t]*$/ && $rawline =~ /^\+[ \t]*\*/ && - $realline > 3) { # Do not warn about the initial copyright comment block after SPDX-License-Identifier + $realline > 3 && # Do not warn about the initial copyright comment block after SPDX-License-Identifier + !ignore_networking_comment_style($realfile)) { WARN("NETWORKING_BLOCK_COMMENT_STYLE", "networking block comments don't use an empty /* line, use /* Comment...\n" . $hereprev); }
Hi Lukas
I have made this patch as I last proposed for how certain files can avoid this warning by adding in a common .ignore file.
For the start, I have added only one directory name ie, "drivers/net/wireless/ralink/rt2x00".
In this time, I also found out that users can also use --ignore NETWORKING_BLOCK_COMMENT_STYLE with checkpatch and it will also result in ignoring this style.
I don't know though which way is more preferred for the MAINTAINERS.
What do you think?
Thanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Re: [Linux-kernel-mentees] [PATCH] checkpatch: ignore files not following NETWORKING_BLOCK_COMMENT_STYLE
In reply to Aditya SrivastavaQuoted 7 lines · Aditya Srivastava
Currently checkpatch.pl gives warning for
NETWORKING_BLOCK_COMMENT_STYLEfor files innet/anddrivers/netwhich do not follow the networking comment style.
But some of these files seem to follow the generic comment style instead of networking style and some rather mixed style of comment.
Quoted 6 lines
For e.g.,
drivers/net/wireless/ralink/rt2x00largely follows generic kernel comment style in spite of being insidedrivers/net.Provide an ignore
file(".networking_block_comment_styles.ignore"), where users can add the files they want to ignore this warning.
I believe that is a really bad design decision here.
Imagine that in one directory, the maintainer wants to adjust ten rules for checkpatch.
Are we going to implement this logic for those ten rules individually? Is the maintainer going to add
None of that is obviously acceptable: 1. because there are better design decisions; 2. because tailoring checkpatch would simply not be worth the cost of having all those individual files laying around in the repository and the complication throughout the whole script.
Also, you already recognized that there is already a feature to ignore or select rules through command-line parameters.
So, how about coming up with a proper .checkpatch config file format and then use the already available feature to configure the checkpatch run?
Lukas
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Re: [Linux-kernel-mentees] [PATCH] checkpatch: ignore files not following NETWORKING_BLOCK_COMMENT_STYLE
Quoted 15 lines · Lukas Bulwahn
Quoted 7 lines · Aditya Srivastava
Currently checkpatch.pl gives warning for
NETWORKING_BLOCK_COMMENT_STYLEfor files innet/anddrivers/netwhich do not follow the networking comment style.
But some of these files seem to follow the generic comment style instead of networking style and some rather mixed style of comment.Quoted 6 lines
For e.g.,
drivers/net/wireless/ralink/rt2x00largely follows generic kernel comment style in spite of being insidedrivers/net.Provide an ignore
file(".networking_block_comment_styles.ignore"), where users can add the files they want to ignore this warning.
Quoted 6 lines
I believe that is a really bad design decision here.
Imagine that in one directory, the maintainer wants to adjust ten rules for checkpatch.
Are we going to implement this logic for those ten rules individually?
No, only the files which are not following the networking comment style, and are inside net/ or drivers/net should be added. The files which are following the networking style already, need not add themselves anywhere.
I feel, normally maintainer will want entire directory to follow the same style. So, they can just add their directory in the .ignore file and it will be ignored from the warning.
Quoted 7 lines
Is the maintainer going to add
None of that is obviously acceptable: 1. because there are better design decisions; 2. because tailoring checkpatch would simply not be worth the cost of having all those individual files laying around in the repository and the complication throughout the whole script.
We have earlier used a similar approach with ".get_maintainer.ignore", where the users add their signature to be ignored from being mentioned by "get_maintainers.pl" script.
We just need one file for our purpose (i.e., not necessarily .ignore file).
Here I have used .ignore file in the root directory, so that it is easier for the maintainer to edit it and add their directory in it.
If not in the root directory(as the files look scattered), perhaps we can place it in script directory as .ignore or .txt file, or maybe just place these directories in the form of a hash in the script itself. But perhaps we'll need a list of directories to ignore and then we'll be able to form the hash.
Quoted 7 lines
Also, you already recognized that there is already a feature to ignore or select rules through command-line parameters.
So, how about coming up with a proper .checkpatch config file format and then use the already available feature to configure the checkpatch run?
The user can add "--ignore NETWORKING_BLOCK_COMMENT_STYLE" line in their .checkpatch.conf file, and checkpatch will then ignore this class of warning.
Somewhat similar to this repository here:
https://github.com/openil/u-boot/blob/master/.checkpatch.conf
But this file(".checkpatch.conf") is ignored by git and is not committed (as different users may require different configurations). So every user will have to add this line in their individual config files.
Also, if the user starts working on a different file, where the networking comment style is followed, he will again have to delete this configuration line from the config file, to get this warning.
I'm not sure, if we can add directory/file specific configuration in .checkpatch.conf. But again, if it was possible, we'll be specifying individual files, and telling them to be avoided, similar to the current approach.
Using any of these solution, if the users ignore this warning, we can perhaps add a fix for NETWORKING_BLOCK_COMMENT_STYLE warning.
What do you think?
Thanks
Aditya
Quoted 2 lines
Lukas
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Re: [Linux-kernel-mentees] [PATCH] checkpatch: ignore files not following NETWORKING_BLOCK_COMMENT_STYLE
Quoted 24 lines · Aditya
Quoted 15 lines · Lukas Bulwahn
Quoted 7 lines · Aditya Srivastava
Currently checkpatch.pl gives warning for
NETWORKING_BLOCK_COMMENT_STYLEfor files innet/anddrivers/netwhich do not follow the networking comment style.
But some of these files seem to follow the generic comment style instead of networking style and some rather mixed style of comment.Quoted 6 lines
For e.g.,
drivers/net/wireless/ralink/rt2x00largely follows generic kernel comment style in spite of being insidedrivers/net.Provide an ignore
file(".networking_block_comment_styles.ignore"), where users can add the files they want to ignore this warning.Quoted 6 lines
I believe that is a really bad design decision here.
Imagine that in one directory, the maintainer wants to adjust ten rules for checkpatch.
Are we going to implement this logic for those ten rules individually?
Quoted 9 lines
No, only the files which are not following the networking comment style, and are inside
net/ordrivers/netshould be added. The files which are following the networking style already, need not add themselves anywhere.I feel, normally maintainer will want entire directory to follow the same style. So, they can just add their directory in the .ignore file and it will be ignored from the warning.
I understand that.
Quoted 8 lines
Quoted 7 lines
Is the maintainer going to add
None of that is obviously acceptable: 1. because there are better design decisions; 2. because tailoring checkpatch would simply not be worth the cost of having all those individual files laying around in the repository and the complication throughout the whole script.
Quoted 4 lines
We have earlier used a similar approach with ".get_maintainer.ignore", where the users add their signature to be ignored from being mentioned by "get_maintainers.pl" script.
Yeah, I know that. It is known to be called the "Christoph file in the root directory", because for years, it only contained Christoph as he was the only person that requested that feature. It is a good example of a terrible feature.
Quoted 6 lines
We just need one file for our purpose (i.e., not necessarily .ignore file).
Here I have used .ignore file in the root directory, so that it is easier for the maintainer to edit it and add their directory in it.
Do you know how many maintainers in this project exist? For the kernel, this what you suggest just does not make sense...
Quoted 6 lines
If not in the root directory(as the files look scattered), perhaps we can place it in script directory as .ignore or .txt file, or maybe just place these directories in the form of a hash in the script itself. But perhaps we'll need a list of directories to ignore and then we'll be able to form the hash.
Which problem are you trying to solve?
Quoted 12 lines
Quoted 7 lines
Also, you already recognized that there is already a feature to ignore or select rules through command-line parameters.
So, how about coming up with a proper .checkpatch config file format and then use the already available feature to configure the checkpatch run?
The user can add "--ignore
NETWORKING_BLOCK_COMMENT_STYLE" line in their .checkpatch.conf file, and checkpatch will then ignore this class of warning.
Quoted 7 lines
Somewhat similar to this repository here:
https://github.com/openil/u-boot/blob/master/.checkpatch.confBut this
file(".checkpatch.conf")is ignored by git and is not committed (as different users may require different configurations). So every user will have to add this line in their individual config files.
I do not get what you say. Isn't this just a configuration question? You configure it to be part of the configuration.
Quoted 9 lines
Also, if the user starts working on a different file, where the networking comment style is followed, he will again have to delete this configuration line from the config file, to get this warning.
I'm not sure, if we can add directory/file specific configuration in .checkpatch.conf. But again, if it was possible, we'll be specifying individual files, and telling them to be avoided, similar to the current approach.
A suitable feature can be implemented.
Think about 100 people modifying the same file at the same time and the problem this brings during merging.
Think about commits moving directories from one place to the other and the effort of keeping this file correct.
I would suggest not to have any "checkpatch.config" at the root level mentioning all the different directories, as you are suggesting.
I suggest instead:
Have multiple ".checkpatch.conf" files in the different directories and let checkpatch determine the relevant rules that need to be applied to a specific file by traverse along the file hierarchy. Generalize the method to work for any checkpatch rule, not just the block_comment_style rule here.
Feel free to submit your current proposal to Joe Perches and lkml if you are convinced of your idea and willing to get more feedback (probably just more rejection). I personally believe this will be rejected by others as well.
Lukas
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees