Bug #5533: segfault when setting xslt parameter in a loop - SaxonC - Saxonica Developer Community

Actions

Send by e-mail Copy link

Bug #5533

closed

segfault when setting xslt parameter in a loop

Added by Lou Burnard over 2 years ago. Updated over 2 years ago.

Status:

Closed

Priority:

Normal

Assignee:

O'Neil Delpratt

Category:

Python

Start date:

2022-05-23

Due date:

% Done:

100%

Estimated time:

Applies to branch:

Fix Committed on Branch:

Fixed in Maintenance Release:

Found in version:

11.3

Fixed in version:

11.4

SaxonC Languages:

SaxonC Platforms:

SaxonC Architecture:

Description

my python code looks like this:

`import saxonc
with saxonc.PySaxonProcessor(license=False) as proc:
    print(proc.version)
    proc.set_configuration_property("xi", "on")
    xdmAtomicval = proc.make_boolean_value(False)
    xsltproc = proc.new_xslt30_processor()
    for LANG in LANGS:
      repoName=repoRoot+LANG
      print("Summarizing repo "+repoName+ " on "+lastUpdate)
      params=' corpus='+LANG + ' lastUpdate='+ lastUpdate
      xsltproc.set_parameter("corpus",proc.make_string_value(LANG)) 
      xsltproc.set_parameter("lastUpdate",proc.make_string_value(lastUpdate))
      src=repoName + "/driver.tei" 
      result = xsltproc.transform_to_string(source_file=src, stylesheet_file=scriptRoot+"summarize.xsl")
      print(result) `

The first time round the loop everything works correctly. But the second iteration (no matter what the input data) always leads to a core dump and the message

`JET RUNTIME HAS DETECTED UNRECOVERABLE ERROR: system exception at 0x00007ff0557820f4
Please, contact the vendor of the application.
Core dump will be written to "/home/lou/Public/Scripts/core"
Extra information about error is saved in the "jet_err_31824.txt" file.`

If I comment out the two invocations of xsltproc.set_parameter everything works fine (or as well as it can without the parameter values)

The jet_err file is attached in the hope that you can spot what is going wrong, or advise me how to pin it down!

Files

Download all files

jet_err_31748.txt (93.4 KB) jet_err_31748.txt	diagnostics from jet	Lou Burnard, 2022-05-23 15:21
summarize2.py (1.59 KB) summarize2.py		Lou Burnard, 2022-05-23 17:55
summarize.xsl (16.5 KB) summarize.xsl		Lou Burnard, 2022-05-23 17:55

Related issues

Actions

Copy link

Updated by O'Neil Delpratt over 2 years ago

Category set to Python
Assignee set to O'Neil Delpratt
Priority changed from Low to Normal
Found in version set to 11.3

This looks like a bug. Thanks for reporting it.

For the make_string_value it looks like python is deleting the object created at the end of the loop thinking that the object is not used again.

As a workaround, for the following code:

      xsltproc.set_parameter("corpus",proc.make_string_value(LANG)) 
      xsltproc.set_parameter("lastUpdate",proc.make_string_value(lastUpdate))

Is it possible to replace it as follows:

      XDM_LANG= proc.make_string_value(LANG)
      xdm_lastUpdate = proc.make_string_value(lastUpdate)
      xsltproc.set_parameter("corpus", XDM_LANG) 
      xsltproc.set_parameter("lastUpdate",xdm_lastUpdate)

I will try to create a repo, but is it possible you can supply the stylesheet please so that I can investigate it further?

Actions

Copy link

Updated by Lou Burnard over 2 years ago

File summarize2.py summarize2.py added
File summarize.xsl summarize.xsl added

Excellent! That work around seems to have been successful. Many thanks for the quick turnround, and good luck hunting down the bug. I am attaching my current version of the python code and the style sheet it invokes, since you asked for it, though they won't make much sense without all the other bits and pieces.

Actions

Copy link

Updated by O'Neil Delpratt over 2 years ago

Alternatively, you can call clear_parameters on the PyXslt30Processor at the end of the for loop. This is the recommended practice.

See the example below:

    for LANG in LANGS:
      repoName=repoRoot+LANG
      print("Summarizing repo "+repoName+ " on "+lastUpdate)
      params=' corpus='+LANG + ' lastUpdate='+ lastUpdate
      xsltproc.set_parameter("corpus",proc.make_string_value(LANG)) 
      xsltproc.set_parameter("lastUpdate",proc.make_string_value(lastUpdate))
      src=repoName + "/driver.tei" 
      result = xsltproc.transform_to_string(source_file=src, stylesheet_file=scriptRoot+"summarize.xsl")
      print(result) 
      xsltproc.clear_parameter()

Actions

Copy link

Updated by O'Neil Delpratt over 2 years ago

Related to Bug #4386: setProperty and setParameter does not replace keys already in the map added

Actions

Copy link

Updated by O'Neil Delpratt over 2 years ago

Status changed from New to Resolved
% Done changed from 0 to 100

Bug fixed in the Xslt30Processor and across other Processors too. There is a regression from bug issue #4386. The check for existing keys in the map failed. This has now been patched and available for the next maintenance release.

The workaround in comment #3 avoids the crash. Also as a speed improvement, I recommend compiling the stylesheet to a PyXsltExecutable object before the for loop, which can then be reused. See below:

    executable = xsltproc.compile_stylesheet(stylesheet_file=scriptRoot+"summarize.xsl")
    for LANG in LANGS:
      repoName=repoRoot+LANG
      print("Summarizing repo "+repoName+ " on "+lastUpdate)
      params=' corpus='+LANG + ' lastUpdate='+ lastUpdate
      executable.set_parameter("corpus",proc.make_string_value(LANG)) 
      executable.set_parameter("lastUpdate",proc.make_string_value(lastUpdate))
      src=repoName + "/driver.tei" 
      result = executable.transform_to_string(source_file=src)
      print(result) 
      executable.clear_parameter()