Buy commercial curl support from WolfSSL. We help you work
out your issues, debug your libcurl applications, use the API, port to new
platforms, add new features and more. With a team lead by the curl founder
himself.
Re: [Help Requested] Test Differences on 64-bit vs. 32-bit NonStop
- Contemporary messages sorted: [ by date ] [ by thread ] [ by subject ] [ by author ] [ by messages with attachments ]
From: Daniel Stenberg via curl-library <curl-library_at_lists.haxx.se>
Date: Sat, 10 Sep 2022 23:37:41 +0200 (CEST)
On Fri, 9 Sep 2022, Randall via curl-library wrote:
> I'm having some difficulties trying to build curl in 64 bits (big endian) on
> NonStop. It works fine, with only 3 minor issues in 32-bit but when
> compiling for 64-bit, a whole bunch of issues come up:
Curious. Is this a new regression? I mean, did previous versions of curl did
any better?
> A sample of the failure is running test 9 in debug. I get the following,
> which seems to indicate a problem writing server.input. I'm at a loss on
> this.
server.input is written by the test server, so when it doesn't exist or is
blank it means the test server didn't receive any protocol bytes to write
there. This would indicate that curl failed before that point in time.
> 9: protocol FAILED!
> There was no content at all in the file log/server.input.
> Server glitch? Total curl failure? Returned: 2011
Returned: 2011 is at least a clue. This is from the test script's:
https://github.com/curl/curl/blob/master/tests/runtests.pl#L4512
I suspect that "$cmdres" gets the value to be > 2000 by a line before that:
https://github.com/curl/curl/blob/master/tests/runtests.pl#L4258
... which would indicate that curl died with a SIGSEGV.
If you rerun it with gdb it probably shows:
$ ./runtests.pl -g 9
(and type 'run' to run it within gdb)
If it indeed does hit a SIGSEGV inside gdb, a stack trace would give us a lot
of clues!
Date: Sat, 10 Sep 2022 23:37:41 +0200 (CEST)
On Fri, 9 Sep 2022, Randall via curl-library wrote:
> I'm having some difficulties trying to build curl in 64 bits (big endian) on
> NonStop. It works fine, with only 3 minor issues in 32-bit but when
> compiling for 64-bit, a whole bunch of issues come up:
Curious. Is this a new regression? I mean, did previous versions of curl did
any better?
> A sample of the failure is running test 9 in debug. I get the following,
> which seems to indicate a problem writing server.input. I'm at a loss on
> this.
server.input is written by the test server, so when it doesn't exist or is
blank it means the test server didn't receive any protocol bytes to write
there. This would indicate that curl failed before that point in time.
> 9: protocol FAILED!
> There was no content at all in the file log/server.input.
> Server glitch? Total curl failure? Returned: 2011
Returned: 2011 is at least a clue. This is from the test script's:
https://github.com/curl/curl/blob/master/tests/runtests.pl#L4512
I suspect that "$cmdres" gets the value to be > 2000 by a line before that:
https://github.com/curl/curl/blob/master/tests/runtests.pl#L4258
... which would indicate that curl died with a SIGSEGV.
If you rerun it with gdb it probably shows:
$ ./runtests.pl -g 9
(and type 'run' to run it within gdb)
If it indeed does hit a SIGSEGV inside gdb, a stack trace would give us a lot
of clues!
-- / daniel.haxx.se | Commercial curl support up to 24x7 is available! | Private help, bug fixes, support, ports, new features | https://curl.se/support.html -- Unsubscribe: https://lists.haxx.se/listinfo/curl-library Etiquette: https://curl.se/mail/etiquette.htmlReceived on 2022-09-10