PuTTY bug assert-line-not-null

Home | Licence | FAQ | Docs | Download | Keys | Links
Mirrors | Updates | Feedback | Changes | Wishlist | Team

We receive occasional reports of PuTTY crashing out with line==NULL in terminal.c:lineptr(). However, we haven't been able to reproduce the problem ourselves.

In recent versions of PuTTY (0.58 and up), the error takes the form of one or more error boxes saying "line==NULL in terminal.c" and giving some diagnostic information, along with an exhortation to contact us.

In versions of PuTTY before snapshot 2005-03-07, the error is instead a failure of an assertion in terminal.c that line != NULL (the precise line number has changed a lot due to the extensive terminal re-engineering since 0.56).

We've had one report since we added diagnostics which indicates that screen doesn't contain enough entries. However, we haven't thought of a plausible mechanism for screen or alt_screen to end up undersized.

(Possibly related, although we've no hard evidence for this: `assert-fail-newsavelines'.)

There seem to be a lot of reports of this error from people using the serial backend in 0.59, which suggests that it can be a symptom of win-handle-double-free (fixed in 0.60).

In fact, in general, we suspect this is a symptom of some sort of memory corruption to which this assertion is particularly sensitive, rather than evidence of a direct logical error.

• [email protected]
  0.56 on WinME and 2K; terminal.c:159
• [email protected]
  0.56 on W2KSP2; terminal.c:159
• [email protected]
  0.56 on Win2K; terminal.c:159
  "disable switching to alternate screen" enabled
• [email protected]
  with diagnostic info
• [email protected]
  0.58 on Win2K+SP4
  

  line==NULL in terminal.c
  lineno=5376 y=65 w=182 h=65
  count(scrollback=00890758)=2
  count(screen=008BE508)=65
  count(alt=00890768)=65 alt_sblines=64
  whichtree=008BE508 treeindex=65
  

• D4EFE24E74186242BCFA48FB1BC47C2802E6681D@ore-mail2.ne.ad.ner.capgemini.com
  0.59, using a serial port
  

  line==NULL in terminal.c
  lineno=4754 y=51 w=153 h=56
  count(scrollback=003F1BF0)=382
  count(screen=003F1C00)=56
  count(alt=00B620C8)=56 alt_sblines=0
  whichtree=003F1C00 treeindex=51
  

• FA37BC94952941AC89BFF98CD9997327@jamesPC
  Also a serial connection
  

  line==NULL in terminal.c
  lineno=4754 y=18 w=89 h=47
  count(scrollback=011E1AA8)=200
  count(screen=011E1AB8)=47
  count(alt=011F9BF0)=47 alt_sblines=0
  whichtree=011E1AB8 treeindex=18
  

• [email protected]
  Using a serial port, so presumably 0.59
  

  linux==NULL in terminal.c
  lineno=4754 y=10 w=80 h=24
  count(scrollback=003A1C60)=200
  count(screen=003A1C70)=24
  count(alt=003A1F90)=24 alt_sblines=0
  whichtree=003A1C70 treeindex=10
  

One user ([email protected]) reported a failure of the same assertion with PuTTY 0.58. However, as their screen is apparently zero rows high (presumably due to their window manager), this pathological case is a problem that's now protected against via `zero-rows-cols-crash'.

I am using bblean blackbox for windows as a shell on windows XP SP2, and whenever I rollup a maximized putty window the following fatal error results:

line==NULL in terminal.c
lineno=4621 y=0 w=93 h=0
count(scrollback=003B0F60)=200
count(screen=00CFA2B8)=0
count(alt=003B0F70)=0 alt_sblines=0
whichtree=00CFA2B8 treeindex=0

Perhaps an unrelated bug that's already been fixed:

• 000d01c4b1c7$ca506090$4296a8c0@nadine
  2004-10-14; terminal.c:662; connecting to SunOS
  reportedly fixed in 2004-10-15, although since other reports indicate the failure is intermittent, perhaps not

Audit trail for this bug.