| 00:19.51 | *** join/#edev NightMonkey (~NightrMon@pdpc/supporter/professional/nightmonkey) |
| 00:49.35 | *** join/#edev jann (~jann@2a00:1a28:1101:63f:1337::948) |
| 02:00.33 | *** join/#edev Belliash (~belliash@asiotec/constitutive/belliash) |
| 02:38.26 | *** join/#edev landley (~landley@70.114.134.56) |
| 02:54.19 | *** join/#edev NightMonkey (~NightrMon@173-164-139-193-SFBA.hfc.comcastbusiness.net) |
| 02:54.20 | *** join/#edev NightMonkey (~NightrMon@pdpc/supporter/professional/nightmonkey) |
| 03:11.41 | *** join/#edev djerome (~djerome@ip68-98-228-41.ph.ph.cox.net) |
| 03:48.49 | *** join/#edev NightMonkey (~NightrMon@pdpc/supporter/professional/nightmonkey) |
| 04:18.03 | *** join/#edev prpplague (~prpplague@107.206.64.184) |
| 04:47.59 | *** join/#edev TheSeven (~quassel@rockbox/developer/TheSeven) |
| 05:09.40 | MonMotha | ok, any ARM people handy? I've got some strange gcc/ld output that I'm trying to debug (getting a usage fault in a veneer to a function in RAM) |
| 05:22.52 | MonMotha | this may actually be a bug... |
| 06:01.47 | *** join/#edev NightMonkey (~NightrMon@pdpc/supporter/professional/nightmonkey) |
| 07:18.30 | *** join/#edev ntfreak_ (~ntfreak@unaffiliated/ntfreak) |
| 09:04.40 | *** join/#edev florian_kc (~fuchs@Maemo/community/contributor/florian) |
| 09:25.46 | *** join/#edev mbuf (~user@182.74.118.98) |
| 09:52.27 | *** join/#edev sirdancealot (~sirdancea@194.228.11.172) |
| 10:16.50 | *** join/#edev devgiant (~devgiant@unaffiliated/devgiant) |
| 12:14.29 | *** join/#edev g1powermac (~g1powerma@unaffiliated/g1powermac) |
| 12:55.52 | *** join/#edev BitEvil (~quassel@tor/regular/SpeedEvil) |
| 13:27.03 | *** join/#edev MonMotha (~monmotha@2001:470:c5f2:1:224:7eff:fedc:893e) |
| 14:50.25 | *** join/#edev Lars_G (~Lars@unaffiliated/lars-g/x-000001) |
| 14:50.34 | *** join/#edev MonMotha (~monmotha@2001:470:c5f2:1:224:7eff:fedc:893e) |
| 15:00.46 | *** join/#edev akp_ (~akp@173-162-162-206-NewEngland.hfc.comcastbusiness.net) |
| 15:08.49 | landley | MonMotha I could take a look? |
| 15:59.45 | *** join/#edev purl (~purl@rikers.org) |
| 15:59.45 | *** topic/#edev is Discussion of embedded systems (Hardware/Software/Linux/Commercial Device Repurposing) : Busybox and Buildroot : http://www.elinux.org : Logs available at ibot.rikers.org |
| 16:07.01 | *** join/#edev kuldeepdhaka (~kuldeepdh@unaffiliated/kuldeepdhaka) |
| 16:17.36 | *** join/#edev espiral (maze@unaffiliated/espiral) |
| 16:34.11 | *** join/#edev prpplague (~danders@rrcs-97-77-26-26.sw.biz.rr.com) |
| 16:34.20 | *** mode/#edev [+o prpplague] by ChanServ |
| 16:52.06 | *** join/#edev ntfreak (~ntfreak@unaffiliated/ntfreak) |
| 17:33.33 | *** join/#edev NightMonkey (~NightrMon@173-164-139-193-SFBA.hfc.comcastbusiness.net) |
| 17:33.33 | *** join/#edev NightMonkey (~NightrMon@pdpc/supporter/professional/nightmonkey) |
| 19:22.52 | *** join/#edev flo_lap (~fuchs@sign-4db6bc9a.pool.mediaWays.net) |
| 19:22.53 | *** join/#edev flo_lap (~fuchs@Maemo/community/contributor/florian) |
| 21:00.26 | *** join/#edev Belliash (~belliash@asiotec/constitutive/belliash) |
| 22:03.01 | *** join/#edev MonMotha (~monmotha@2001:470:c5f2:1:224:7eff:fedc:893e) |
| 22:03.17 | MonMotha | landley: you around? |
| 22:51.43 | calculus | landley: did I tell you I had liquid nitrogen prepared ice cream last year? |
| 22:52.01 | calculus | also liquid nitrogen drenched graham crackers |
| 22:55.04 | MonMotha | landley: the issue seems to be that ld is generating a veneer for a long call to a function in ram (from a function in flash), and that veneer wants to put the MCU into ARM (not Thumb) mode. Of course, that doesn't fly on a Cortex-M4. Everything works in a couple older compilers, which may make it a regression if indeed it is a bug, but I'm not 100% sure yet that I've got all my settings right. |
| 22:57.22 | calculus | I would also suggest reaching out to ssvb (not sure if he frequents #edev, but he does show up in #gentoo-embedded) |
| 22:58.34 | SpeedEvil | Liquid oxygen soaked crackers are more fun. |
| 22:58.57 | roxfan | i think i remember seeing this bug somewhere before |
| 22:59.02 | roxfan | maybe on some ML |
| 23:00.09 | roxfan | http://sourceware.org/ml/binutils/2011-07/msg00174.html |
| 23:01.34 | roxfan | how are you peforming the call? show the code |
| 23:02.23 | *** join/#edev prpplague (~danders@rrcs-97-77-26-26.sw.biz.rr.com) |
| 23:02.24 | *** mode/#edev [+o prpplague] by ChanServ |
| 23:04.08 | MonMotha | roxfan: it's just a straight C function call to the function, which is statically defined int he same file. I use a GCC attribute to force that function into a special section (.ram) which the linker places in RAM, and crt0 initializes at the same time as .data is copied. It is also marked with the "never inline" attribute. |
| 23:04.21 | MonMotha | The function has to be in RAM to work around a hardware errata |
| 23:04.33 | roxfan | hmm |
| 23:07.05 | roxfan | maybe run readelf -s on the image and check if the target symbol address has bit 0 set |
| 23:07.08 | roxfan | just to be sure |
| 23:07.40 | roxfan | oh btw you aren't passing -mthumb-interwork or similar? |
| 23:10.16 | roxfan | http://sourceware.org/bugzilla/show_bug.cgi?id=15302 #2 looks like your case |
| 23:13.21 | roxfan | pokes MonMotha |
| 23:14.07 | MonMotha | roxfan: I am not using -mthumb-interwork, but I also am not using -mno-thumb-interwork |
| 23:15.08 | roxfan | see last link |
| 23:27.10 | MonMotha | roxfan: looks like that could be it. I'm running 2.24, so I may be affected. |
| 23:27.40 | MonMotha | I guess I can try that patch and see what happens |
| 23:28.02 | MonMotha | my calls are happening from C so the stub name is different, but I wouldn't be surprised if the same logic goes into generating the veneer for a C function |
| 23:29.45 | MonMotha | that comment says he's getting a hard fault. I get a usage fault, which make sense (the UFSR says I'm trying to operate in ARM mode). I guess maybe the commenter is just not handling the usage fault properly and it's escalating? |
| 23:30.05 | roxfan | probably |
| 23:31.40 | MonMotha | I wonder when that regression happened |
| 23:54.00 | landley | MonMotha: just got back. |
| 23:54.30 | MonMotha | landley: I was discussing what was happening with roxfan just above. Looks like it may in fact be a known (and already patched upstream) GNU ld bug |
| 23:54.31 | landley | Alas, I'm not familiar with what current gcc is doing. Poking at llvm a bit, though. :) |
| 23:54.40 | landley | calculus: liquid nitrogen is cool. :P |
| 23:55.25 | landley | Indeed, caught up now. :) |
| 23:55.52 | MonMotha | I guess I need to patch and rebuild my binutils |
| 23:56.13 | landley | I've come to the conclusion that upgrading FSF software is always a mistake. |
| 23:56.21 | landley | Sometimes it's the lesser of two evils, but it's never a good thing. |
| 23:56.58 | MonMotha | heh, yeah, for now I just moved back to my known-working toolchain |