[ https://issues.apache.org/jira/browse/HADOOP-12301?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Kengo Seki updated HADOOP-12301: -------------------------------- Description: Some test-patch plugins have problems about counting the diff lines. 1. ruby-lint.sh counts only lines which have five or more fields: {code} diffPostpatch=$(${AWK} -F: 'BEGIN {sum=0} 4<NF {sum+=1} END {print sum}' "${PATCH_DIR}/diff-patch-ruby-lint.txt") if [[ ${diffPostpatch} -gt 0 ]] ; then # shellcheck disable=SC2016 numPrepatch=$(${AWK} -F: 'BEGIN {sum=0} 4<NF {sum+=1} END {print sum}' "${PATCH_DIR}/branch-ruby-lint-result.txt") # shellcheck disable=SC2016 numPostpatch=$(${AWK} -F: 'BEGIN {sum=0} 4<NF {sum+=1} END {print sum}' "${PATCH_DIR}/patch-ruby-lint-result.txt") {code} This is because the diff result can contain multiple lines for one issue. For example: {code} [sekikn@mobile hadoop]$ cat /tmp/test-patch-hbase/1149/diff-patch-rubocop.txt bin/draining_servers.rb:165:1: C: Do not introduce global variables. $FOO ^^^^ bin/draining_servers.rb:166:1: C: Do not introduce global variables. $BAR ^^^^ {code} Checking the number of fields is intended to skip the second, third, fifth and sixth lines above. But four or more colons can appear in the source code itself, so more reliable way is needed. 2. pylint.sh has the same problem. In addition, I removed awk's '-F:' option by mistake on HADOOP-12286. It must report a wrong number for now. 3. rubocop.sh should also follow ruby-lint.sh. Or, change its formatter to "emacs" or "simple" style (currently "clang"). They are parser-friendly because they outputs one line by one issue, so we can count issues by 'wc -l' like perlcritic.sh. https://github.com/bbatsov/rubocop#emacs-style-formatter was: Some test-patch plugins have problems about counting the diff lines. 1. ruby-lint.sh counts only lines which have five or more fields: {code} diffPostpatch=$(${AWK} -F: 'BEGIN {sum=0} 4<NF {sum+=1} END {print sum}' "${PATCH_DIR}/diff-patch-ruby-lint.txt") if [[ ${diffPostpatch} -gt 0 ]] ; then # shellcheck disable=SC2016 numPrepatch=$(${AWK} -F: 'BEGIN {sum=0} 4<NF {sum+=1} END {print sum}' "${PATCH_DIR}/branch-ruby-lint-result.txt") # shellcheck disable=SC2016 numPostpatch=$(${AWK} -F: 'BEGIN {sum=0} 4<NF {sum+=1} END {print sum}' "${PATCH_DIR}/patch-ruby-lint-result.txt") {code} This is because the diff result can contain multiple lines for one issue. For example: {code} [sekikn@mobile hadoop]$ cat /tmp/test-patch-hbase/1149/diff-patch-rubocop.txt bin/draining_servers.rb:165:1: C: Do not introduce global variables. $FOO ^^^^ bin/draining_servers.rb:166:1: C: Do not introduce global variables. $BAR ^^^^ {code} Checking the number of fields is intended to skip the second, third, fifth and sixth lines above. But four or more colons can appear in the source code itself, so more reliable way is needed. 2. pylint.sh has the same problem. In addition, I removed awk's '-F:' option by mistake on HADOOP-12286. It must report a wrong number for now. 3. On the other hand, rubocop.sh counts the lines in the same way as ruby-lint.sh, but rubocop seems to report one issue by one line. So it'd better to count by 'wc -l', in the same as perlcritic.sh. > Fix some test-patch plugins to count the diff lines correctly > ------------------------------------------------------------- > > Key: HADOOP-12301 > URL: https://issues.apache.org/jira/browse/HADOOP-12301 > Project: Hadoop Common > Issue Type: Sub-task > Components: yetus > Affects Versions: HADOOP-12111 > Reporter: Kengo Seki > > Some test-patch plugins have problems about counting the diff lines. > 1. ruby-lint.sh counts only lines which have five or more fields: > {code} > diffPostpatch=$(${AWK} -F: 'BEGIN {sum=0} 4<NF {sum+=1} END {print sum}' > "${PATCH_DIR}/diff-patch-ruby-lint.txt") > if [[ ${diffPostpatch} -gt 0 ]] ; then > # shellcheck disable=SC2016 > numPrepatch=$(${AWK} -F: 'BEGIN {sum=0} 4<NF {sum+=1} END {print sum}' > "${PATCH_DIR}/branch-ruby-lint-result.txt") > # shellcheck disable=SC2016 > numPostpatch=$(${AWK} -F: 'BEGIN {sum=0} 4<NF {sum+=1} END {print sum}' > "${PATCH_DIR}/patch-ruby-lint-result.txt") > {code} > This is because the diff result can contain multiple lines for one issue. For > example: > {code} > [sekikn@mobile hadoop]$ cat /tmp/test-patch-hbase/1149/diff-patch-rubocop.txt > bin/draining_servers.rb:165:1: C: Do not introduce global variables. > $FOO > ^^^^ > bin/draining_servers.rb:166:1: C: Do not introduce global variables. > $BAR > ^^^^ > {code} > Checking the number of fields is intended to skip the second, third, fifth > and sixth lines above. But four or more colons can appear in the source code > itself, so more reliable way is needed. > 2. pylint.sh has the same problem. In addition, I removed awk's '-F:' option > by mistake on HADOOP-12286. It must report a wrong number for now. > 3. rubocop.sh should also follow ruby-lint.sh. Or, change its formatter to > "emacs" or "simple" style (currently "clang"). They are parser-friendly > because they outputs one line by one issue, so we can count issues by 'wc -l' > like perlcritic.sh. https://github.com/bbatsov/rubocop#emacs-style-formatter -- This message was sent by Atlassian JIRA (v6.3.4#6332)