HPCC-18308 eclcmd should pass eclcc options as debug values #10443
Conversation
|
https://track.hpccsystems.com/browse/HPCC-18308 |
|
@ghalliday Please sanity check this |
ecl/eclcmd/eclcmd_core.cpp
Outdated
| @@ -148,6 +148,8 @@ bool doDeploy(EclCmdWithEclTarget &cmd, IClientWsWorkunits *client, const char * | |||
| expandDefintionsAsDebugValues(cmd.definitions, cmd.debugValues); | |||
| if (cmd.debugValues.length()) | |||
| { | |||
| IArrayOf<IEspNamedValue> debugValues; | |||
| @@ -254,7 +254,8 @@ class EclccCompileThread : implements IPooledThread, implements IErrorReporter, | |||
| { | |||
| //Allow eclcc-xx-<n> so that multiple values can be passed through for the same named debug symbol | |||
| const char * start = option + (*option=='-' ? 1 : 6); | |||
| const char * dash = strrchr(start, '-'); // position of second dash, if present | |||
| const char * finger = (*start=='-') ? start+1 : start; //support leading double dash | |||
There was a problem hiding this comment.
I don't understand this change - can you elaborate?
There was a problem hiding this comment.
@ghalliday It's slightly confusing, but the key is that when the option is appended to the eclcc command line the stripped '-' is restored. I was trying not to refactor the code more than I had to as this is a last minute PR.
eclccCmd.appendf(" -%s=%s", optName.get(), value);
Some eclcc options actually have two dashes. "--nostdinc" for example. So this change preserves the second dash in the option name "-nostdinc" and then both dashes end up on the command line.
|
@afishbeck I may be misunderstanding, but it looks like an option -f-x is converted to eclcc-x, but would then be matched in eclccserver as option x. I would have expected it to be converted to eclcc--x (then the double minus code would make sense.) |
"ecl run <target> <file> -f-<option>" Allows arbitrary eclcc options to be passed to the eclcc command line. Those options should also be passed as workunit debug values formatted as "eclcc-<option>". Signed-off-by: Anthony Fishbeck <anthony.fishbeck@lexisnexis.com>
afishbeck
force-pushed
the
eclccDebugValueOptions
branch
from
b0158dc
to
8e3fb66
Compare
|
@ghalliday Removed unused variable. Responded to your other comment. |
|
Automated Smoketest: ✅
Install hpccsystems-platform-community_6.4.2-rc3.el7.x86_64.rpm Unit tests result:
Regression test result:
HPCC Stop: OK |
|
@richardkchapman looks good. I'll let you decide which build it goes in. |
|
I have a concern that --fastparse and --nostdinc should be being applied to the creation of the archive when ecl command calls out to a local eclcc, but don't necessarily apply to the parsing at the remote end - so while it's ok to fix the -f- processing as you have here, we really need --fastparse and --nostdinc to be also handled explicitly by all ecl commands. I also suspect we may need to push this out to 6.4.4. There is a workaround of creating the archive independently I think. |
|
@richardkchapman I could easily add those as explicit eclcmd options. My guess would have been that --nostdinc had no effect on the backend since the depencies were already gathered into the archive at the client side? I can't seem to find --fastparse. |
|
As mentioned in the meeting, --fastparse was a typo, it's --fastsyntax. |
richardkchapman
changed the base branch from
candidate-6.4.2
to
candidate-6.4.4
richardkchapman
changed the base branch from
candidate-6.4.4
to
candidate-6.4.6
"ecl run -f-"
Allows arbitrary eclcc options to be passed to the eclcc command line.
Those options should also be passed as workunit debug values formatted
as "eclcc-".
Signed-off-by: Anthony Fishbeck anthony.fishbeck@lexisnexis.com
Type of change:
Checklist:
Testing: