*** tpb has joined #symbiflow | 00:00 | |
*** mzyn22 has joined #symbiflow | 02:53 | |
*** mzyn22 has left #symbiflow | 02:56 | |
*** citypw has joined #symbiflow | 03:07 | |
*** Bertl_zZ is now known as Bertl | 06:10 | |
*** OmniMancer has joined #symbiflow | 06:20 | |
*** _whitelogger has quit IRC | 10:57 | |
*** _whitelogger has joined #symbiflow | 10:59 | |
*** OmniMancer has quit IRC | 11:31 | |
*** Bertl is now known as Bertl_oO | 12:08 | |
*** freemint has quit IRC | 12:12 | |
*** freemint has joined #symbiflow | 13:12 | |
*** craigo has quit IRC | 13:13 | |
*** freemint has quit IRC | 14:13 | |
*** freeemint has joined #symbiflow | 14:13 | |
*** lromor[m] has quit IRC | 14:19 | |
*** xobs has quit IRC | 14:19 | |
*** synaption[m] has quit IRC | 14:19 | |
*** mrhat2010[m] has quit IRC | 14:19 | |
*** hzeller[m] has quit IRC | 14:19 | |
*** zeigren has quit IRC | 14:19 | |
*** nrossi has quit IRC | 14:19 | |
*** galv[m] has quit IRC | 14:19 | |
*** alexhw[m] has quit IRC | 14:19 | |
*** alexhw has quit IRC | 14:33 | |
*** alexhw has joined #symbiflow | 15:17 | |
*** alexhw has quit IRC | 15:20 | |
*** citypw has quit IRC | 16:01 | |
*** Bertl_oO is now known as Bertl | 16:03 | |
*** nrossi has joined #symbiflow | 16:33 | |
*** alexhw[m] has joined #symbiflow | 17:13 | |
*** hzeller[m] has joined #symbiflow | 17:13 | |
*** zeigren has joined #symbiflow | 17:13 | |
*** synaption[m] has joined #symbiflow | 17:13 | |
*** xobs has joined #symbiflow | 17:13 | |
*** galv[m] has joined #symbiflow | 17:13 | |
*** lromor[m] has joined #symbiflow | 17:13 | |
*** mrhat2010[m] has joined #symbiflow | 17:13 | |
*** freeemint has quit IRC | 17:35 | |
*** freeemint has joined #symbiflow | 17:35 | |
litghost | mkurc: Can you make a wip/ branch version on https://github.com/SymbiFlow/vtr-verilog-to-routing of https://github.com/verilog-to-routing/vtr-verilog-to-routing/pull/993 ? | 17:57 |
---|---|---|
tpb | Title: GitHub - SymbiFlow/vtr-verilog-to-routing: SymbiFlow WIP changes for Verilog to Routing -- Open Source CAD Flow for FPGA Research (at github.com) | 17:57 |
litghost | mkurc: We cannot predict when Kevin or another VTR dev will be able to merge that PR | 17:58 |
*** freeemint has quit IRC | 18:34 | |
sf-slack | <mkurc> @litghost Sure, I'll address your review and then create the wip/ branch. | 19:08 |
*** freeemint has joined #symbiflow | 19:19 | |
*** freeemint has quit IRC | 19:44 | |
*** freeemint has joined #symbiflow | 19:44 | |
*** freeemint has quit IRC | 19:55 | |
*** freeemint has joined #symbiflow | 19:56 | |
*** adjtm_ has joined #symbiflow | 20:34 | |
*** adjtm has quit IRC | 20:37 | |
*** Bertl is now known as Bertl_zZ | 20:45 | |
*** craigo has joined #symbiflow | 20:55 | |
*** freeemint has quit IRC | 21:36 | |
hackerfoo | litghost: I haven't had time to reproduce the lxml crash yet, but I found this: https://bugs.launchpad.net/lxml/+bug/1570388 | 21:52 |
tpb | Title: Bug #1570388 “Serialisation error when writing a large file (> 2...” : Bugs : lxml (at bugs.launchpad.net) | 21:52 |
litghost | If that is the issue, it was reported 3 years ago, and got absolutely no updates | 21:54 |
litghost | So :( | 21:54 |
litghost | However it does seem like it has a good chance of matching our issue | 21:56 |
litghost | I'm trying their replication now to see if it fails in the same way | 21:57 |
litghost | I'm working on getting usan/asan/etc to run against libxml2/libxslt and see what (if anything) crops up | 21:58 |
hackerfoo | lxml uses libxml2: https://bugzilla.gnome.org/show_bug.cgi?id=796192 | 21:58 |
tpb | Title: Bug 796192 libxml2, xml file reader issue when file size is greater than 2GB (at bugzilla.gnome.org) | 21:58 |
litghost | hackerfoo: Yes, which makes finding the bug more "exciting" | 21:58 |
litghost | hackerfoo: The bug could exist in lxml, libxml2, or be an exciting combination !!! | 21:59 |
hackerfoo | It could still be in our code, too. | 22:02 |
litghost | Nope! Because https://bugs.launchpad.net/lxml/+bug/1570388 is exactly our bug | 22:07 |
tpb | Title: Bug #1570388 “Serialisation error when writing a large file (> 2...” : Bugs : lxml (at bugs.launchpad.net) | 22:07 |
litghost | Like exact same symptom and everything | 22:07 |
litghost | Now it is possible that they (openpyxl) have the same bug we do, but that feels unlikely | 22:09 |
hackerfoo | Well, now we have code to reproduce the bug. | 22:09 |
litghost | Kind of, it's still intermixed with openpyxl, so it is no better per say than what we have | 22:12 |
litghost | A straight forward large loop didn't replicate | 22:12 |
*** freeemint has joined #symbiflow | 22:17 | |
duck2 | did we upgrade to a >2.5gb xml file? :D | 22:21 |
litghost | ~3.8 GB :D | 22:22 |
*** freeemint has quit IRC | 22:22 | |
litghost | I found the libxml2 bug, and it super stupid | 22:56 |
hackerfoo | What was it? | 23:01 |
litghost | int written ... written += bytes_write_last_call ... if (written < 0) error | 23:01 |
litghost | e.g. what we expected | 23:02 |
litghost | libxml2 is totally misusing int (versus size_t/ssize_t) | 23:02 |
hackerfoo | I wonder if Clang or GCC has a warning for that (casting size_t/ssize_t to int.) | 23:03 |
litghost | hackerfoo: It does not, but "-fsanitize=undefined" with clang produced: | 23:06 |
litghost | xmlIO.c:3416:19: runtime error: signed integer overflow: 2147481777 + 4000 cannot be represented in type 'int' | 23:06 |
litghost | e.g. overflowing written | 23:06 |
hackerfoo | So can we use the whole 50T now? | 23:07 |
duck2 | oooh | 23:08 |
litghost | I'm working on a patch, but I believe there was never actually a problem in the output, except that the written counter overflowed | 23:08 |
litghost | And this is why size_t and ssize_t should always be used over int | 23:09 |
duck2 | so it's why it occurs when the file is >2gb | 23:09 |
litghost | We suspected it was a 32-bit overflow issue, but the question was where | 23:09 |
*** freeemint has joined #symbiflow | 23:33 | |
*** alainmarcel has joined #symbiflow | 23:57 |
Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!