help hcs_ gets record_quota_overflow for accounts with default sized process directories

[Eric Swenson]

If you invoke the "help hcs_" command, you get the following:

help hcs_

Error:  record_quota_overflow condition by info_seg_allocate_$iSect|1343
(>system_library_standard>bound_info_rtns_)
referencing !BBBKQGgxdlPqMf.temp.0471|112234 (in process dir)

r 08:52 1.276 804 level 2

rl
r 08:52 0.097 0

gq [pd]
quota = 1000; used = 196
r 08:52 0.058 0

We probably didn't notice this earlier, when testing the new "help", because both Gary and I have higher-than-default process directory quotas.

[Eric Swenson]

In fact, this is a bit worse than I expected. I just did a new_proc (in a process with only 1000 records of quota) and then "help hcs_". Not only did I get the record_quota_overflow, but it terminated my process.

[Gary Dixon]

Clearly hcs_.info is the largest subroutine info segment with 102 entry points. "help hcs_" requires about 1300 records of temp segs in the process directory to parse all of its entry points (as measured by a recent test).

">obs>old_help hcs_" requires about 70 records to get a list of entry points.

I remember now having to add a special code path in old_help (when it was "the help") to deal with very large info subroutine info segments. Looks like such might be needed again.

[Gary Dixon]

Note that the existing info_seg_ code (and the temp segment manager code in general) carefully truncate temp segments when they are returned to the pool for re-use by another program. help_ always returns temp segments to the pool at the end of each help command.

[Gary Dixon]

Note that the verify_info command also shares this problem when asked to verify contents of a large info segment (such as hcs_.info).

[Eric Swenson]

This is essentially the same issue as https://multics-trac.swenson.org/ticket/281.

​https://s3.amazonaws.com/eswenson-multics/public/mcrs/MCR10122.pdf

[Eric Swenson]

Fixed with MR12.8-1039 install.