1748073 Members
5425 Online
108758 Solutions
New Discussion юеВ

Re: "strings core"

 
ranganath ramachandra
Esteemed Contributor

"strings core"

please do not suggest this as a way to debug coredump issues.
 
--
ranga
[i work for hpe]

Accept or Kudo

8 REPLIES 8
Dennis Handly
Acclaimed Contributor

Re: "strings core"

Yes, I always say that. Was there a particular thread you wanted to reply to, instead of starting a new thread?

You do know you can "fix" that by making all error message strings be in read only memory, which won't be in the core file.
Laurent Menase
Honored Contributor

Re: "strings core"

don't agree,
On some cases like memory leak, - and only in that case
strings core | tail -10000 | head -1000
- -10000 and -1000 are not deterministic values-
can sometimes help to identify a pattern.

It can gives sometimes tracks to explore, even if in any case it will not give directly the answer.
ranganath ramachandra
Esteemed Contributor

Re: "strings core"

ok, i should have specifically asked not to suggest this to users who are clueless about core files, if only because it can easily lead them off track, as they don't know what they are looking for.

even if a user is knowledgeable enough to make sense of the stuff they get out of "strings core", they are very likely to be good enough with a debugger anyway.

a debugger would provide enough tools to see exactly what (and where) one is looking for.

as for shifting error messages into read-only text, that has to be done in all runtime components. but even when that is done, i still don't see how "strings core" could be better than a debugger given the potential wild-goose-chases.

no use making the user print strings, confuse themselves and open the problem up to a hundred interpretations by everyone who wants to help.

i started a new thread to avoid having to repeat this in every other thread. now i (or you) just have to link this one into all those threads :)
 
--
ranga
[i work for hpe]

Accept or Kudo

Dennis Handly
Acclaimed Contributor

Re: "strings core"

>asked not to suggest this to users who are clueless about core files, if only because it can easily lead them off track, as they don't know what they are looking for.

Exactly.

>as for shifting error messages into read-only text, that has to be done in all runtime components. but even when that is done, i still don't see how "strings core" could be better than a debugger given the potential wild-goose-chases.

There are several reasons:
1) There will be a performance benefit.
2) If the error strings aren't there, nobody will suggest using strings(1).
3) If a string does show up, it may actually be important.

>no use making the user print strings, confuse themselves and open the problem up to a hundred interpretations by everyone who wants to help.

Right.
Laurent Menase
Honored Contributor

Re: "strings core"

I agree on the fact that debugging just need the executable & libs, the corefile, a debugger, - the source is usually usefull too- and skills.

So in any case there is no need to ask for a strings core if you are trying a remote debug.
You gain time when you ask for corefiles, libs, and executables.


strings core is just a trick which may be used, but not as a first approach in any case.
Hein van den Heuvel
Honored Contributor

Re: "strings core"

ranganath,

Please indicate which topic made you write this topic.

I tried to find with a simple google: +site:itrc.hp.com "strings core"
It found many hits, including this, but nothhing otherwise recent.

Google did report something intersting, which I had not really noticed before.
It suggested:

"IT Resource Center forums - core files - This thread has been closed 10 posts - 8 authors - Last post: Mar 16, 2007
:
forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1107726 "

Does that "10-posts 8 authors - last post" not suggest that Google 'understands' itrc topics?

The 'source' for the page in question does have:
<meta name="replyauthors" content="Sunny Jaisinghani, Adisuria Wangsadinata, Reshma Malusare, Eric Jacklin, Matti Kurkela, Dennis Handly, Johnny Damtoft, prash" />

That could explain the '8 authors', but it does not have an (obvious) reply count of last reply date (GMT vs US?)

Just curious.... and not on topic. Sorry.

Hein.

ranganath ramachandra
Esteemed Contributor

Re: "strings core"

dennis,

there's no question on why error messages should be moved to read-only text.

hein,

this is not about one particular thread but the ITRC search shows up these recent ones.

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1272520
http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1262594
http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1255611

and the stuff that you posted about google search results is interesting. it seems that only some threads have this information, though.
 
--
ranga
[i work for hpe]

Accept or Kudo

ranganath ramachandra
Esteemed Contributor

Re: "strings core"

discussed enough
 
--
ranga
[i work for hpe]

Accept or Kudo