[xslt] Override of attributes from attribute set

[Bob Kline <bkline rksystems com>]

We've been struggling with an elusive bug which had us baffled for a while. The XSL/T spec says that if you attach an attribute directly to a literal element, which also pulls in that attribute with an xsl:use-attribute-sets attribute, then the attribute specified on the literal result element will override the value set in the attribute sets. This has worked fine for years, and then all of a sudden table borders started disappearing with no explanation, and the same code would produce different output on the lower tiers than what we got on the production server. I was able to determine that libxslt version 1.1.29 was handling the attributes correctly, but 1.1.30 was not, nor was 1.1.33. I was preparing a repro case so I could file a bug and then I discovered that the latest version of the Python lxml package is built and distributed with different versions of libxslt, depending which platform OS is targeted, and when I tested on my MacBook, the repro case didn't reproduce the failure. The version of libxslt on that machine was 1.1.34, which appears to have fixed the regression. So the only remaining puzzle is: why is there nothing in the bug tracker (would have saved a bunch of time and work)? Did this really get discovered and fixed without a ticket? Or did the bug get introduced _and_ fixed accidentally, without anyone noticing either change? Or is there a bug report in the tracking DB that I'm just not seeing? If the bug really did come and go without anyone being aware of it, and you would like to record it (if nothing else, adding to the regression test suite), I'd be happy to provide my repo case. If there's an appropriately titled bug report, and my old eyes and/or brain are just too dim to have found it, please accept my embarrassed apologies. :-)

Thanks!

--

Bob Kline

https://www.rksystems.com

mailto:bkline rksystems com