gen-snmp-subdriver.sh hangs

[darkglade]

Hi All,

Been trying to use gen-snmp-subdriver.sh to generate a stub for my Eaton/Liebert GXT4 UPS with an IS-WEBCARD management interface, I thought the SNMP walk was just taking a really long time for some reason but in checking the process listing (and embroidering the script with some temporary debug statements) it's actually these lines from the file (490-491);

${EGREP} "^[ ${TABCHAR}]?\$" | ${GREP} "End of MIB" | ${GREP} -v "${NUMWALKFILE}" > "${TMP_NUMWALKFILE}"
${EGREP} "^[ ${TABCHAR}]?\$" | ${GREP} "End of MIB" | ${GREP} -v "${STRWALKFILE}" > "${TMP_STRWALKFILE}"

Unless I'm completely missing something (or my shell is doing something very different to what the committer's was) I can't see how that's ever going to do anything BUT hang? There's no file input to any of the grep commands in the chain until the last one so there's never anything on STDIN so it never actually does anything?

Seems like this change was specifically to avoid the use of -e so I presume that's not well supported, I think that the following should work and satisfy the criteria (certainly works fine for me, though my snmpwalks are NOT generating any blank lines or "End of MIB" lines)

${EGREP} -v "^[ ${TABCHAR}]?\$" "${NUMWALKFILE}" | ${GREP} -v "End of MIB" > "${TMP_NUMWALKFILE}"
${EGREP} -v "^[ ${TABCHAR}]?\$" "${STRWALKFILE}" | ${GREP} -v "End of MIB" > "${TMP_STRWALKFILE}"

Happy to raise a PR for it, but just wanted to sanity check first in case I'm completely missing something.

[jimklimov]

Thanks for the catch, this seems to be a regression introduced after v2.8.4 release with PR #3107 for issue #3099. I guess there was a lot of semi-automated change and not enough review for code not hit during CI builds...

:; git diff v2.8.4 scripts/subdriver/gen-snmp-subdriver.sh
...

 # remove blank and "End of MIB" lines
-grep -E -e "^[[:space:]]?$" -e "End of MIB" -v "${NUMWALKFILE}" > "${TMP_NUMWALKFILE}"
-grep -E -e "^[[:space:]]?$" -e "End of MIB" -v "${STRWALKFILE}" > "${TMP_STRWALKFILE}"
+TABCHAR="`printf '\t'`"
+${EGREP} "^[ ${TABCHAR}]?\$" | ${GREP} "End of MIB" | ${GREP} -v "${NUMWALKFILE}" > "${TMP_NUMWALKFILE}"
+${EGREP} "^[ ${TABCHAR}]?\$" | ${GREP} "End of MIB" | ${GREP} -v "${STRWALKFILE}" > "${TMP_STRWALKFILE}"
 NUMWALKFILE="${TMP_NUMWALKFILE}"
 STRWALKFILE="${TMP_STRWALKFILE}"

Your suggested change seems reasonable; alas for portability's sake it can not be simplified to a group with parentheses (that trick is not in Basic Regular Expressions vocabulary).

[jimklimov]

Revised, seems that was the only such grep -e miss in that PR.

[jimklimov]

Thanks, I see you had a commit ready - made a PR from it :)

[darkglade]

No worries, I was planning to submit it myself but got distracted before I finished reviewing the project's requirements for PRs ;)